GSkillPacks

`./publish.sh --dry-run --tag experimental --preid experimental prerelease`

`./scripts/skill-mirror-parity-audit.sh`

`./scripts/skill-pack-routing-audit.sh`

`./scripts/skill-versions.sh --missing`

`## Invoke With YAML` is described as optional clean-context routing metadata for the next invocation.

`agent_routing` YAML is not the only human-facing UAT command.

`agentic-skills` no longer has the active copied alignment page or index card, but still keeps `research/youtube-owner-analytics-platform.md`.

`alignment/index.html` links the new page exactly once with a `2026-07-01` date.

`bash scripts/skill-archive-audit.sh --strict`

`bash scripts/skill-mirror-parity-audit.sh`

`business-research` is the canonical Business AFPS pack name; `business-discovery` appears only as compatibility/alias history.

`CHANGELOG.md` has an empty `[Unreleased]` placeholder and a `## [0.1.19] - 2026-07-02` release-prep section.

`CLAUDE.md` states BIP is never a blocker, required approval gate, required review step, or downstream-routing prerequisite outside `idea-scope-brief`.

`cleanup --all --global` is rejected.

`cleanup --global --dry-run` scans projects under the fake/user home, not the command cwd, and still previews legacy user-home skill removals.

`cleanup --global --reinstall-base` uses the user-home project scan root for migration.

`cleanup`, `cleanup --all`, and `cleanup --all --dry-run` scan projects from the command cwd.

`competitive-analysis`

`create-alignment-page` can rely on packaged scaffold support when source checkout docs/scripts are unavailable.

`dead-code`, `slim-audit`, `debug`, and `hygiene` remain present but are routed narrowly against `quality-sweep`, `investigate`, and `reconcile-dev-docs`.

`docs/decks.md` lists VARD and ORD chains with `*-traction`.

`enterprise-icp`

`experimental` points to `.6` for both packages while `latest` stays at `0.1.21`.

`flow-tree.schema.json` uses `schema_version: v0.5`, keeps the six-step route unchanged, and makes `platform_fit` optional/additive.

`GeorgeQLe/youtube-owner-analytics` exists as a private repo with no GitHub Actions added by this task.

`git diff --check --cached`

`git diff --check`

`gskp alignment pages scaffold <skill> <topic> --out alignment/<skill>-<topic>.html` writes a safe repo-relative page.

`gskp interrogation pages scaffold <skill> <round> <branch> --out interrogation/<skill>-r<round>-<branch>.html` writes a safe repo-relative page.

`logic-wiring` and `consolidate-prototypes` both identify `uat` as provided by `product-testing` when unavailable.

`node --test packages/skillpacks/test/briefing-slides-audit.test.mjs`

`node --test packages/skillpacks/test/lifecycle.test.mjs packages/skillpacks/test/project-config.test.mjs packages/skillpacks/test/compatibility.test.mjs`

`node --test packages/skillpacks/test/lifecycle.test.mjs` (76/76)

`node --test packages/skillpacks/test/pack-normalization.test.mjs`

`node --test packages/skillpacks/test/package-boundary.test.mjs packages/skillpacks/test/publish-recovery.test.mjs`

`node scripts/audit-briefing-slides.mjs`

`node scripts/audit-task-docs.mjs`

`node scripts/skill-convention-bundle-audit.mjs`

`node scripts/upgrade-alignment-page.mjs --check`

`node scripts/upgrade-interrogation-page.mjs --check`

`npm --workspace packages/skillpacks run build:check`

`npm --workspace packages/skillpacks run build:check` (passed after temporarily generating a stable-lane manifest; final source restored to canary)

`npm --workspace packages/skillpacks run build:manifest:check`

`npm --workspace packages/skillpacks run test:node`

`npm --workspace packages/skillpacks run test:node` (217/217)

`npm run exports:check`

`npm view @glexcorp/gskp dist-tags --json`

`npm view @glexcorp/gskp@0.1.22-experimental.5 version`

`npm view skillpacks dist-tags --json`

`npm view skillpacks@0.1.22-experimental.5 version`

`npm_config_cache=/tmp/skillpacks-npm-cache npm --workspace packages/skillpacks run test:node`

`npx -y --package @glexcorp/gskp@experimental -- skillpacks list`

`npx -y --package skillpacks@experimental -- skillpacks install code-quality`

`npx -y --package skillpacks@experimental -- skillpacks install quality-sweep`

`npx -y --package skillpacks@experimental -- skillpacks list`

`npx -y --prefer-online --package @glexcorp/gskp@experimental --cache /tmp/skillpacks-npm-cache -- skillpacks --version`

`npx -y --prefer-online --package @glexcorp/gskp@experimental --cache /tmp/skillpacks-npm-cache -- skillpacks list`

`npx -y --prefer-online --package skillpacks@experimental --cache /tmp/skillpacks-npm-cache -- skillpacks --version`

`npx -y --prefer-online --package skillpacks@experimental --cache /tmp/skillpacks-npm-cache -- skillpacks list`

`npx skillpacks cleanup` removes legacy managed globals and BIP config from discovered projects.

`npx skillpacks uninstall-global` still works as a deprecated alias.

`packages/skillpacks/dist/skillpacks-manifest.json` package metadata is `{ "version": "0.1.22-experimental.1", "release_lane": "canary" }`.

`packages/skillpacks/dist/skillpacks-manifest.json` package metadata is `{ "version": "0.1.22-experimental.5", "release_lane": "canary" }`.

`packages/skillpacks/package.json` remains at `0.1.18`.

`packages/skillpacks/package.json` version is `0.1.22-experimental.1`.

`packages/skillpacks/package.json` version is `0.1.22-experimental.5`.

`platform_fit.candidates[]` covers the broad platform set, fit/status values, evidence basis, platform-specific risk fields, required probe, and recommendation buckets.

`pnpm --dir tests exec vitest run --project layer1 layer1/interrogation-confidence-gate.test.ts`

`pnpm --dir tests test layer1/bip-removed.test.ts`

`pnpm --dir tests test layer1/install.test.ts`

`pnpm --dir tests test layer1/skill-install-routing-audit.test.ts`

`pnpm --dir tests test layer1/skills-showcase-pack-coverage.test.ts`

`prototype_build_item.platform_probe` allows explicitly non-visual platform probes without requiring `ui_experiment_id`, while ordinary UI build items still link to a UI experiment.

`prototype` and `create-ui-experiment` are absent from active skill directories, pack skills, catalog cards, and route targets, with replacement guidance to `logic-wiring` and `build-ui-screens`.

`README.md` states the repo owns the YouTube owner analytics wrapper work and points the next step at `$investigate` with `tasks/next-step.yaml`.

`refresh --all --dry-run` reports updates without mutation; `refresh --all` completes affected projects as `ok` rather than `failed`.

`scripts/validate-skills-catalog-export.sh`

`ship`, `ship-end`, and Codex `exec` produce an issue, non-primary branch, pushed commits, and ready PR when GitHub is available, while merge remains a separate confirmed action.

`ship`/`ship-end` prompt only in terminal output when `alignment.build_in_public` is absent and `alignment.bip_prompt_dismissed !== true`.

`ship`/`ship-end` treat `alignment.build_in_public: false` as "do not ask and do not generate BIP."

`SKILLPACKS_PACKAGE_LANE=canary npm --workspace packages/skillpacks run build:check`

`SKILLPACKS_PACKAGE_LANE=canary npm_config_cache=/tmp/skillpacks-npm-cache npm --workspace packages/skillpacks run verify:package`

`skillpacks@experimental` and `@glexcorp/gskp@experimental` both resolve to `0.1.22-experimental.1`.

`skillpacks@experimental` and `@glexcorp/gskp@experimental` both resolve to `0.1.22-experimental.5`.

`skillpacks@latest` and `@glexcorp/gskp@latest` both remain `0.1.21`.

`Staying in this session is allowed` is removed or made explicitly exceptional in generated design-tree-loop bundles.

`tasks/next-step.yaml` contains the exact prior owner analytics investigation prompt content that the transfer plan identifies as the user-provided next-step YAML source.

`ui-interview` review pages explicitly warn that words like missing/reject/retry/revision in approving copy must not affect readiness without blocking metadata.

`ux-variations` and `ui-interview` no longer rely only on implicit "first pending" order when recommending child branches.

`vard-ship` and `ord-ship` route to traction before AFPS.

A second refresh is idempotent, and `refresh --all` preserves per-project failure isolation.

Active BIP audit accepts and enforces the dedicated `alignment/bip/` directory, and rejects stale top-level or old checkpoint shapes.

Active Claude `SKILL.md` files do not contain user-facing dollar invocation commands where a `/skill` command is intended.

Active Codex `SKILL.md` files do not contain user-facing slash invocation commands where a `$skill` command is intended.

Active non-archived skill/convention text has no BIP blocker language except the allowed `idea-scope-brief` gate and normal non-BIP alignment approval gates.

Active skill stubs can continue to reference `Follow the shared shipping contract convention in CLAUDE.md`.

Actual refactors require version, archive, and changelog handling per skill versioning rules.

Add `--allow-dirty-tree` parsing, usage text, tracked/untracked dirty summary output, and release-impacting path classification.

Add `--prefer-online` to the `npx` package execution path while retaining the isolated npm cache.

Add `### Shipping Contract Convention` to the canonical Claude and AGENTS provision blocks.

Add `bin/skillpacks.mjs`.

Add `CHANGELOG.md` entries for v0.1 in each skill directory.

Add `cleanup` as the primary CLI command while preserving `uninstall-global` as a compatibility alias.

Add `doctor --fix --agent-docs [--dry-run]` for marker-bounded generated block replacement only.

Add `doctor --fix` to clean generated skill-root drift without touching agent docs.

Add `game-afps` to `packages/skillpacks/scripts/build-skillpacks-manifest.mjs`.

Add `interrogation-page` to parent and framework `required_conventions`.

Add `packages/skillpacks/package.json` with the publishable package metadata.

Add `packages/skillpacks/scripts/build-skillpacks-manifest.mjs`.

Add `packs/base/PACK.md` and establish `packs/base/{claude,codex}` as the canonical base-pack shape.

Add `packs/product-design/codex/user-flow-map/` and `packs/product-design/claude/user-flow-map/`.

Add `Report-First Approval Gate`, `Staged Research Workflow`, and `Evidence And Feedback Handling`.

Add `scripts/catalog/*.mjs` helpers for frontmatter parsing, pack/skill scanning, archive discovery, benchmark report discovery, and content hashing.

Add `scripts/lib/` to package build staging.

Add `scripts/lib/` to package publish files.

Add `scripts/lib/` to the npm `files` allowlist.

Add `scripts/lib/` to the skillpacks package build copy boundary.

Add `scripts/researchish-skill-lifecycle-audit.mjs`.

Add `skillpacks install-deck <deck>` and `skillpacks install-deck business-afps --full`.

Add `skillpacks list --json` using the manifest.

Add `src/cli/run-pack-script.mjs` or equivalent dispatcher.

Add `ui-interview`-specific alignment guidance requiring a visible interview-stage explainer.

Add `user-flow-map` to the alignment generator and regenerate alignment-page stubs.

Add a `0.1.19` release-prep section to `CHANGELOG.md` while leaving package source version at `0.1.18`.

Add a `Deck Fit Handoff` rule that reads canonical deck metadata and `.agents/project.json` `saved_decks` / `decks`.

Add a canary-specific publishing steps script.

Add a canonical GitHub delivery contract and a validator that rejects active direct-to-primary mutation instructions outside release/deploy exceptions.

Add a clear review-stage answer: `requirements-only` review now, full interview only when the skill is run interactively or rerun without `--requirements-only`.

Add a convention rule that clicking the empty deck stage around the visible slide advances to the next slide.

Add a dated history entry and ship manifest.

Add a dated Product Design & Spec card for the walkthrough page.

Add a lesson for future addenda that introduce new decisions.

Add a lesson preventing future alignment-page chart clipping.

Add a lesson preventing research synthesis before alignment-scope approval.

Add a Node-owned `skillpacks init` command that installs global-scope skills as project-local base skills under `.claude/skills` and `.codex/skills`.

Add a non-following path inspection helper that distinguishes an absent path from a dangling symlink and propagates non-`ENOENT` errors.

Add a package build script that stages only package-owned code, canonical skill sources, required install scripts, and selected docs into `packages/skillpacks/build/`.

Add a package test that asserts Game AFPS resolves to the `game` pack and carries the right registry tags.

Add a Playwright spike test or document a dependency/install blocker before shipping.

Add a required deck-installation approval gate covering COA A/B/C behavior.

Add a root `scripts/init-agentic-skills.sh` wrapper that delegates to the bundled `init-agentic-skills` launcher.

Add active roadmap and todo entries before substantive audit work.

Add active roadmap and todo entries before substantive edits.

Add active roadmap and todo tracking before implementation.

Add active-slide YAML and all-feedback YAML compilation/copy controls to the sidebar.

Add alignment-page metadata, TTS includes, and index entries for the two untracked alignment pages.

Add an audit helper that detects active final artifact approval gates from gate/question containers carrying final-artifact, canonical-artifact, or artifact-approval metadata without depending only on `data-required="true"`.

Add an enforceable BIP checkpoint contract to the canonical alignment convention and regenerate generated bundles from that source.

Add an explicit post-investigation user-approval gate before any implementation path.

Add assemble-ready carve-out to **§5 Progress Handoff Block** — `docs/design-tree-loop-convention.md` (~lines 454–481) — at the assemble-ready stop where the deliverable's alignment page is in `review`, the block must not foreground "repeat this command" or a `## Invoke With YAML` payload; it foregrounds "Review required" and presents only the compiled response YAML. Reserve repeat-command framing for setup + per-unit stops.

Add benchmark coverage metadata or an explicit deferred coverage path.

Add BIP config cleanup for discovered projects, including dry-run reporting.

Add bottom-bar `Feedback` control, right-side sidebar, close button, and Escape close behavior.

Add bounded retry handling around published-package smoke commands in `packages/skillpacks/scripts/verify-published-package.sh`.

Add canonical `docs/briefing-slides-convention.md`.

Add canonical convention text for required-gate slide root status and red/green border behavior.

Add concise guidance that low/medium post-canonical corrections can route to `research-amend`, while high/systemic changes still route to targeted framework/synthesis/full reruns.

Add current npm registry preflight findings for `skillpacks` and `agentic-skills`.

Add dependency checks for `bash` and write-command `jq`.

Add equivalent source-checkout `scripts/pack.sh` install/refresh/doctor behavior.

Add factual reconciliation evidence to history and reconciliation report.

Add focused CLI and fixture audit tests.

Add focused coverage for a mocked real `patch` run where the first `npm publish` exits nonzero after web-auth style output.

Add focused coverage for an auth-preflight failure in a real `patch` run restoring source package metadata.

Add focused layer-1 contract coverage and run the requested tests/audits.

Add focused layer-1 coverage for rename, optional feeder/amendment role, `source_artifacts[]`, structured board contract, and route exclusion.

Add focused layer-1 regression coverage for metadata-based approval status and the positive `No decision-critical coverage is missing.` fixture.

Add focused layer1 audit fixtures for passing/failing initial BIP channel-selection pages and selected-channel BIP draft pages.

Add focused layer1 deck graduation routing coverage.

Add focused layer1 routing coverage for both mirror syntaxes.

Add focused lifecycle tests for base init, refresh, and prune expectations.

Add focused parser, lifecycle, and compatibility coverage.

Add focused product-design flow-tree regression assertions.

Add focused publish-script tests for canary prerelease publishing, guardrail failures, stable non-`latest` opt-in, dry-run behavior, and canary `--current` recovery.

Add focused regression coverage for marker-only source drift in normal refresh and `refresh --all --dry-run`.

Add focused regression coverage for web-auth preflight, publish interruption rollback, repeated interrupt during cleanup, token/legacy-auth pass-through, and existing recovery paths.

Add focused regression coverage or static contract checks proving enabled BIP produces post-suggestion instructions.

Add focused regression fixtures/tests for BIP-enabled missing checkpoint failure, linked sibling BIP pass, BIP-disabled pass, Stage 1 no-fail, convention/bundle propagation, and final handoff routing.

Add focused tests and run verification.

Add focused tests for cleanup aliasing, BIP removal, dry-run behavior, and command help.

Add focused tests for default tracked dirty blocking, allowed non-release dirty paths with untracked files, rejected release-impacting dirty paths, supported flag ordering, and unknown flag rejection.

Add focused tests for package assets, single-skill install, pack install, refresh restore, doctor stale/missing reports, and shell install/refresh restore.

Add focused verifier and publish recovery regression tests.

Add fresh-audience context requirements to BIP generation: explain project/work context, define jargon, state why the work matters, and tie each post angle to an understandable public hook.

Add Game AFPS to `docs/decks.md`.

Add Game AFPS to `docs/skillpacks-npm-distribution.md`.

Add interrupt-safe rollback handling for `INT`, `TERM`, and `HUP`, preserving bumped files only after the first publish succeeds.

Add layer1 coverage for staged markers, skip-list bundles, non-research `_working` misuse, and stable JSON categories/counts.

Add lifecycle coverage for read-only doctor, fix-only cleanup, dry-run agent-doc diff, backup/write migration, text preservation, and refusal cases.

Add lifecycle coverage proving `skillpacks install exec` installs the individual `exec` skill and does not enable `exec-loop`.

Add lifecycle coverage that generated `CLAUDE.md` and `AGENTS.md` contain the shared shipping contract pointer and rules.

Add lifecycle regression coverage and update `CHANGELOG.md`.

Add lifecycle regressions for dangling current, legacy, unmanaged, pinned, dry-run, multi-project, removal, pruning, doctor, and idempotent behavior.

Add managed metadata/hash comparison so doctor can report stale or missing convention docs.

Add manifest `--check` validation for existing paths, version fields, deck pack references, package-list fields, and registry-tag fields.

Add migration guidance for users moving from a local clone to `npx skillpacks`.

Add mirrored `take-inspiration` skills with initial changelogs and required bundled conventions.

Add mirrored Codex and Claude `create-briefing-slides` base skill source directories with version `v0.0`.

Add npm script entry points for package and root invocation.

Add or extend package-owned tests that pin the release-readiness documentation language.

Add or update focused layer1 checks that pin the fallback contract.

Add or update focused tests that would catch the dirty metadata behavior.

Add or update routing notes for consolidation candidates without deleting them.

Add package-boundary, CLI, and fixture audit coverage.

Add package-owned Node helpers for `.agents/project.json` reads and writes.

Add package-owned tests that run with `PATH` emptied to prove no shell/JQ dependency.

Add packaged template assets under `packages/skillpacks/assets/templates/`.

Add parent State G stage-zero interrogation before State E framework selection.

Add Playwright as local implementation proof tooling if dependency installation succeeds.

Add prominent canary-channel warning language to `scripts/publish-canary-steps.sh`.

Add regression coverage for Codex and Claude skill contracts, then run focused tests and archive/diff checks.

Add release-channel guardrails so prerelease versions cannot publish to `latest`, stable `latest` publishes cannot carry prerelease versions, and both package names always share version and dist-tag.

Add review notes to `tasks/todo.md`.

Add review-gate precedence clause to the **Pre-approval stop** — `docs/alignment-page-convention.md` (~line 123) — when a chunked assemble stop's next action is the binding review gate (page in `review`), the Pre-approval stop governs over any chunk Progress Handoff repeat-command framing; present exactly one YAML (the page's compiled response YAML). Write generically for all chunked design skills.

Add root `package.json` for `skillpacks`.

Add scaffold templates and CLI command resolution.

Add shared handoff verification classifications and conservative fallback rule to `docs/design-tree-loop-convention.md`.

Add shared resolver/stub contract to the convention registry and bundle audit.

Add shell installer coverage for nested framework archives and lifecycle coverage for `positioning/frameworks/category-design`.

Add staged/tarball regression assertions for `scripts/lib/collapsing-fill-audit.mjs`.

Add static audit diagnostics and fixture coverage for missing required-gate slide markers, status, border styles, and status updater logic.

Add strict scaffold commands for `alignment pages scaffold` and `interrogation pages scaffold`.

Add targeted executable checks for manifest generation and deck install resolution.

Add targeted lifecycle/CLI regression tests.

Add the active roadmap and todo entries before implementation.

Add the downstream portability addendum covering macOS stock Bash compatibility and Bash 3.2 failures in inventory scripts.

Add the pack-install script surface follow-up covering `pack.sh`, no-hot-reload/install visibility guidance, downstream/manual runnable readiness, and the `scripts/init-agentic-skills.sh` path mismatch.

Add the plain-text clarification that deterministic repeatable primitives should become scripts or script-backed utilities, with `compile-central-alignment` as the primary extraction candidate.

Add the shared handoff verification classifications and conservative fallback rule to `docs/design-tree-loop-convention.md`.

Add the smallest route spike that proves shallow `pushState`, `usePathname`, `popstate`, and `/deck/[slug]` hard-load behavior.

Add the standard `## Alignment Page` stub.

Add troubleshooting for npm package semver vs skill-level pinning.

Add UAT post-readiness guard language for result-log categories, manual-evidence capture while any built result is `Not run`, and single-variant MVP override only after explicit user scope choice.

Add unit coverage for `exec`, `exec-loop`, exact pack title, rejected aliases, rejected fuzzy names, and exact skill names.

Add URL ledger behavior for per-video `research/youtube/data/<video-id>/prelaunch/video-url-record.json` records and aggregate `research/youtube/data/video-url-index.jsonl`.

Add visible amendment notes to the three active pages.

Add workspace recognition for `apps/skills-showcase` and `packages/skillpacks`.

Added lifecycle contract coverage and completed archive, parity, catalog, task-doc, alignment, and runtime consistency verification; the full layer-one run passed 2,511/2,532 tests with 21 unrelated failures in untouched research/design/provisioning contracts.

Added the Codex-only Sol/Luna/Terra convention and risk-based topology.

After the cleanup commit, the old repo dirty status contains only the unrelated package metadata edits that existed before this task.

Agent-specific reload/session guidance points at Codex for Codex skills and Claude Code for Claude skills.

Alignment and interrogation conventions describe dense pages as canonical backup/reference surfaces when a briefing deck exists and preserve dense-page open fallback when the briefing convention asset is absent.

Alignment and task-doc audits pass, and the diff contains no unrelated package metadata changes.

All phase tests pass.

All-approve compiles to complete gate and response statuses with `approval_status: ready-for-agent-review`.

Amend `briefing-slides/create-briefing-slides.html` with guarded `A`/`D` hotkeys and empty-stage click navigation.

Amend the dogfood deck with concise slide content, reference chips, visible update marking, and no hidden-overflow substitute for content fitting.

Any changed active `SKILL.md` files follow archive/version/changelog requirements.

Any patch reduces repeated validation work without weakening package boundary allow/deny assertions, release lane assertions, manifest freshness checks, or publish/recovery behavior.

Apply current/legacy managed-link ownership consistently across refresh planning and mutation, doctor, prune, remove, active installs, and pinned installs.

Archive `alignment/idea-scope-brief-npm-distribution.html`.

Archive `alignment/idea-scope-brief-skills-showcase.html`.

Archive `alignment/workflow-design-three-pipelines.html`.

Archive `global/codex/idea-scope-brief/SKILL.md` with `scripts/skill-archive.sh`.

Archive `ui-interview` v0.29, bump Codex and Claude mirrors to v0.30, and update changelogs.

Archive active `ux-variations` skill contracts before behavior changes.

Archive active alias skill directories for Claude and Codex under `archive/deprecated-skills/2026-07-product-design-aliases/`.

Archive affected active `SKILL.md` files before behavior/output changes.

Archive and bump affected `ship-end` skill mirrors if the contract changes behavior.

Archive and bump both `provision-agentic-config` skill variants.

Archive and bump both active `ui-interview` skill mirrors to `v0.30`.

Archive and bump existing skill mirrors before editing their `SKILL.md` contracts.

Archive and bump mirrored `upgrade-interrogation-pages` skills to `v0.2`, preserving/creating hidden agent payloads during upgrades.

Archive and bump versions for every active skill whose behavior/output guidance changes, and update `CHANGELOG.md`.

Archive and rename `design-inspirations` to `brainstorm-inspirations` in both mirrors.

Archive and version active `ux-variations` skill contracts before behavior edits.

Archive and version affected `SKILL.md` files:

Archive and version affected active `SKILL.md` files:

Archive and version affected active `SKILL.md` files.

Archive and version affected Codex skills, add contract coverage, refresh generated outputs, and run the full layer-one suite (2,511/2,532 passed; 21 failures are in untouched research/design/provisioning contracts).

Archive and version the active Codex/Claude `uat`, `logic-wiring`, and `consolidate-prototypes` skill contracts.

Archive and version the four active product-design skills before behavior changes.

Archive and version the four active source shipping skills before behavior changes.

Archive both active mirrors at `v0.4`, bump active `SKILL.md` files to `v0.5`, and update changelogs.

Archive changed active `SKILL.md` files before version bumps.

Archive current `competitive-analysis` parent mirrors and four framework subskill mirrors before editing.

Archive current `v0.3` skill files with `scripts/skill-archive.sh`, bump both active mirrors to `v0.4`, and update matching changelogs.

Archive every behavior-changing skill version, bump versions, and update mirrored changelogs before editing active contracts.

Archive the current `briefing-slides/create-briefing-slides.html` deck before replacement.

Archive the current `investigate` skill version and bump the active skill version for the behavior change.

Archive the current active page before replacement.

Archive the current Codex and Claude `idea-scope-brief` skills and bump active versions to `v0.23`.

Archive the current dogfood deck before amendment.

Archive the current review page and non-canonical working packet under `docs/history/archive/2026-06-11/142531/`.

Archive the prior review alignment page and convert the active page to `confirmed`.

Archive, bump, and update changed mirrored skills and changelogs.

Artifact-only skills are explicitly out of scope for this todo unless later promoted by audit.

Assert `scripts/lib/collapsing-fill-audit.mjs` exists in staged package output and packed tarball coverage.

Assert npm metadata against the package name/version/license.

Attempt to open the generated page with the repository helper.

Audit `tasks/roadmap.md`, `tasks/todo.md`, advisory task files, recent git history, and the task-doc audit.

Audit active YouTube skill data needs, especially `--owner-analytics <path>`, `research/youtube/data/`, and report evidence expectations.

Automated audits fail on regression to direct-primary mutation shipping or on missing mirrored/versioned subskill artifacts.

Avoid skill/pack catalog generated-data changes when tracked `SKILL.md` / `PACK.md` metadata does not change.

Baseline timings identify the slowest reproducible local package validation bottleneck.

Behaviorally important instructions remain present in either the skill or the referenced document.

BIP candidates must include fresh-audience context, jargon expansion, public-facing significance, source basis, claim safety, risk level, publish precheck, loaded convention path, and recommendation status.

BIP commands and kanban surfaces remain removed/hibernated from active discovery.

BIP output remains generated only when enabled or explicitly invoked, and it is read-only help/review content.

Blocking, clarification, and unresolved negative section-feedback paths compile to `approval_status: not-approved`.

Both mirrors of `porter-five-forces`, `swot`, `strategic-group-map`, and `feature-pricing-matrix` declare `interrogation-page`.

Both npm package names publish `0.1.22-experimental.6` on `experimental`; both `latest` tags remain `0.1.21`.

Both parent mirrors declare `required_conventions: [alignment-page, interrogation-page]`.

Both parent mirrors run stage-zero interrogation before framework selection on cold start.

Both published-package smoke matrices passed for `skillpacks` and `@glexcorp/gskp` at `0.1.22-experimental.6`.

Branch routing is deterministic and explicitly tied to journey sequence or a recorded user override.

Brief covers manual, daily channel, and recent-upload refresh scenarios plus expected failure handling.

Brief includes an API capability matrix with sourced claims for OAuth, quotas, metrics, delays, retention, and Reporting API tradeoffs.

Brief makes a build/no-build recommendation for a local-first CLI wrapper.

Brief names the current repo/skill data needs and how normalized files satisfy them without changing `youtube-video-audit`.

Brief specifies raw and normalized evidence layout under `research/youtube/data/`, including metric provenance and explicit gaps.

Build `alignment/devtool-docs-audit-docs-freshness.html`.

Build `alignment/ui-interview-skill-execution-handoff.html` in `review` state with the full working packet rendered inline and required gates.

Build a `review` alignment page with evidence matrix, assumptions, gaps, and approval/feedback controls.

Build and verify `briefing-slides/github-delivery-architecture.html` as the visual approval surface.

Build one convention-doc registry/helper from tracked canonical docs:

Build or run a scriptable parser over the selected history scope to count skill invocation patterns by source and actor.

Build or run an audit for repeated normalized paragraphs within active skills and against direct Markdown references.

Bump `version:` from `v0.15` to `v0.16`.

Bump changed skill versions and update each skill `CHANGELOG.md`.

Bump each active scan skill from `version: v0.0` to `version: v0.1`.

Bump parent mirrors to `v0.29` and framework mirrors to `v0.11`.

Canary `build:check`, `verify:package`, publish dry run, task-doc audit, and diff hygiene passed.

Canary package contents include briefing-slide convention assets and briefing-first alignment/interrogation assets.

Canary package staging still emits canary-only briefing-slide assets and manifest entries.

Canonical extracted `CLAUDE.md` and `AGENTS.md` blocks contain the shipping contract heading, pointer text, next-step routing, commit/push, clean tracked tree, and safety-exception rules.

Capture baseline timings for `npm --workspace packages/skillpacks run test:node`, canary `build:check`, canary `verify:package`, and the slowest individual Node test files surfaced by output.

Capture prompt history and task tracking for the invoked skill work.

Capture prompt history for the active skill update.

Capture prompt history for the active skill-building workflows.

Capture the `$reconcile-dev-docs fix tasks` invocation prompt.

Capture the implementation plan in `tasks/roadmap.md` and `tasks/todo.md`.

Capture the investigation prompt and current task state.

Capture the visible `$analyze-sessions` prompt in `prompts/analyze-sessions/`.

Capture the visible `$exec p1 and p2 again` invocation.

Capture the visible `$investigate` invocation prompt and promote this investigation into `tasks/roadmap.md` and `tasks/todo.md`.

Capture the visible `$investigate` invocation under `prompts/investigate/`.

Capture the visible `$session-triage` invocation prompt and verify the current dirty release-state files.

Capture the visible `$ship-end` invocation under `prompts/ship-end/`.

Capture the visible `devtool-docs-audit` invocation in `prompts/devtool-docs-audit/`.

Capture the visible `devtool-docs-audit` invocation under `prompts/devtool-docs-audit/`.

Capture the visible `exec` invocation in `prompts/exec/`.

Capture the visible `exec` invocation prompt and promote this implementation into `tasks/roadmap.md` and `tasks/todo.md`.

Capture the visible `exec` invocation prompt and promote this recovery fix into `tasks/roadmap.md` and `tasks/todo.md`.

Capture the visible `exec` invocation under `prompts/exec/`.

Capture the visible `idea-scope-brief` approval YAML under `prompts/idea-scope-brief/`.

Capture the visible `investigate` invocation prompt and promote this implementation into `tasks/roadmap.md` and `tasks/todo.md`.

Capture the visible `investigate` invocation prompt and promote this investigation into `tasks/roadmap.md` and `tasks/todo.md`.

Capture the visible `investigate` invocation under `prompts/investigate/`.

Capture the visible `ship` invocation prompt and preserve the starting git status.

Capture the visible `targeted-skill-builder` invocation under `prompts/targeted-skill-builder/`.

Capture the visible Browser verification invocation under `prompts/browser/`.

Capture the visible Computer Use verification invocation under `prompts/computer-use/`.

Capture the visible investigation prompt under `prompts/investigate/`.

Capture the visible invocation under `prompts/exec/` and `prompts/animation-design-planner/`.

Capture the visible invocation under `prompts/exec/`.

Capture the visible invocation under `prompts/pack/`.

Capture the visible invocation under `prompts/targeted-skill-builder/`.

Capture the visible prompt under `prompts/investigate/`.

Capture the visible prompt under `prompts/ui-interview/`.

Capture the visible skill invocation under `prompts/create-briefing-slides/`.

Capture the visible skill update prompt in `prompts/skill-creator/`.

Capture the visible skill-creation invocation under `prompts/skill-creator/`.

Capture the visible skill-update invocation prompt and promote the implementation into `tasks/roadmap.md` and `tasks/todo.md`.

Capture the visible user invocation under `prompts/create-agentic-skill/`.

Capture this implementation plan in `tasks/roadmap.md` and `tasks/todo.md`.

Capture visible `$ui-interview --requirements-only skill-execution-handoff` handoff prompt history.

Capture visible `$user-flow-map skills-showcase skill-execution-handoff` prompt history.

Carry corrected persona wording into the proposed spec and interview log.

Central alignment index guidance treats `alignment/bip/*.html` as BIP review/help pages without making them look like new primary alignment pages.

Change the publish rollback boundary from "first publish command invoked" to "first package publish completed successfully."

Changed `SKILL.md` files are archived and versioned with changelog entries.

Changed `SKILL.md` files are archived, versioned, and have changelog entries.

Check canonical workflow/routing docs against active skill contracts and recent rename/routing work.

Check docs that mention missing or moved paths, stale commands, retired skill names, or generated-file conventions.

Check git status and package metadata.

Check local release posture, published versions, and current package metadata.

Check README/AGENTS/setup docs against current scripts, pack commands, init flow, and skill layout.

Check whether any existing skill already owns user-flow mapping.

Check working tree cleanliness and tracked alignment pages.

Choose the minimal universal architecture: three mirrored base subskills (`github-issue`, `github-branch`, `github-pr`) plus compatibility wrappers.

Chunked `ux-variations` Progress Handoff Blocks no longer require or emit `Exact next command:`.

Clarify `new` marker behavior with dated meta text.

Clarify inherited interrogation plus multi-select approval satisfies inline framework pre-research gates.

Classify core routing skills separately from repeated pack-availability guard updates.

Classify each in-scope skill into exactly one category: `staged-research`, `alignment-document`, `direct-utility`, or `misclassified`.

Claude handoffs include `npx skillpacks install product-testing`, then `/uat --variant-evaluation`.

Clickable UI experiment buildout has a dedicated owner separate from default `ui-interview` behavior.

Codex and Claude state-model `SKILL.md` mirrors carry the same fresh/clean-context cue.

Codex handoffs include `npx skillpacks install product-testing`, then `$uat --variant-evaluation`, and mention a fresh Codex CLI session fallback.

Commit `Release skillpacks 0.1.22-experimental.1`, tag `skillpacks-v0.1.22-experimental.1`, and push `master` plus the tag.

Commit and push intended audit artifacts without staging unrelated pre-existing work.

Commit and push intended changes in each repo, then update task docs. **Reconciliation:** remotes already populated/pushed; this task-doc reconciliation closes out the split.

Commit and push intended changes on `master` while preserving unrelated dirty work.

Commit and push intended changes on `master`.

Commit and push intended changes on the primary branch while preserving unrelated dirty work.

Commit and push intended changes on the primary branch without staging unrelated YouTube work.

Commit and push intended changes on the primary branch.

Commit and push intended changes only, leaving unrelated worktree changes untouched.

Commit and push intended changes only.

Commit and push intended changes.

Commit and push intended files on `master`.

Commit and push intended files on the primary branch.

Commit and push intended Phase 4 changes only.

Commit and push intended Phase 5 changes only.

Commit and push intended readiness artifacts on the primary branch if validation allows.

Commit and push intended source, report, task, and generated-data changes on the primary branch.

Commit and push intended tracked changes if validation passes.

Commit and push intended tracked changes, without touching unrelated local edits.

Commit and push on `master`.

Commit and push only prompt/task/history/manifest files.

Commit and push the intended BIP cleanup boundary on the primary branch while preserving unrelated dirty work.

Commit and push the intended changelog/task/prompt audit boundary on `master` so `git status` is clean.

Commit and push the intended changes only.

Commit and push the intended changes.

Commit and push the intended tracked changes while leaving unrelated work untouched.

Commit and push the intended tracked changes.

Commit and push the new repo seed, then commit and push the `agentic-skills` cleanup separately.

Commit and push the source fix to `master` (`5ab7edbbb`).

Commit source + regenerated bundles + manifest together on `master` and push — commit `a81db4dc7`, pushed `a5098efa4..a81db4dc7`; scoped to these files, pre-existing untracked `.agents/skillpacks/` left untouched (concurrent-session caution).

Commit the convention and regenerated files together.

Commit, tag, and push post-publish metadata; verify dist-tags and published smoke commands.

Compare alignment-page convention requirements across canonical docs, generated bundles, active HTML pages, and audit scripts.

Compare CLI command names and npm/package usage across root docs, package docs, skill docs, and alignment pages.

Compare deck/pack names and counts across docs, manifests, generated maps, and package tests.

Compare related YouTube pack skills for routing and naming overlap.

Compare the new approval against the previously shipped roadmap prompt.

Compare workflow/git/shipping instructions across `AGENTS.md`, `CLAUDE.md`, and repo docs.

Compatibility aliases are described as aliases, not active packs, while old installs still route to replacements.

Compile `CHANGELOG.md` `0.1.14` release notes from the net `v0.1.13..HEAD` change set, keeping package-level notes distinct from individual skill changelogs.

Compilers derive `approval_status` from outcome metadata plus unresolved section feedback, not labels, values, prose, regexes, or substrings.

Confirm `alignment/index.html` is untracked and should remain a local convenience artifact.

Confirm `packages/skillpacks/build/package.json` is `0.1.1`.

Confirm `skillpacks@0.1.14` and `@glexcorp/gskp@0.1.14` are both absent from the npm registry.

Confirm clean starting state on `master` and npm latest versions `0.1.19` for both package names.

Confirm current docs define `npx skillpacks install <pack-or-skill>` and `npx skillpacks install-deck <deck>` as the package install routes.

Confirm existing source-checkout routes remain valid.

Confirm focused regression coverage blocks duplicate chunked continuation routing.

Confirm generated bundles are exact with `node scripts/upgrade-alignment-page.mjs --check`.

Confirm generated count terms still match current generated data.

Confirm it is already tracked in `7ac9ebc3 docs: audit skillpacks cli routing gaps`.

Confirm it is prompt-history bookkeeping only and contains no obvious secret.

Confirm license/package metadata choice based on repository files.

Confirm no other SKILL.md hard-codes the assemble-stop framing — checked `state-model`, `ui-interview`, `user-flow-map` (both codex+claude). None hard-code emitting the full repeat-command Progress Handoff Block **at** the assemble+approve review-page stop: `state-model` §4 says "Stop for compiled YAML", `ui-interview`/`user-flow-map` say "build the one alignment page … on approval …". Their Progress Handoff Block applies to setup/per-unit/synthesis-**ready** handoffs (into the assemble session), not the review-gate stop. ux-variations was unique in its step-8 "emit the required Progress Handoff Block" at that stop. Siblings rely on the shared convention; the §5 edit (regenerated into all `DESIGN-TREE-LOOP.md` bundles) covers them. No additional SKILL.md edits needed.

Confirm npm dist-tag parity for both package names.

Confirm npm registry metadata for both package names and the `experimental` dist-tag.

Confirm repository context and working-tree state.

Confirm the compiled gate list includes the new deck-installation gate.

Confirm the desired local directory layout and whether missing GitHub repos should be created. **Decision:** repos already exist (none to create); single-checkout layout, no local sibling clones.

Confirm the dry-run reports `skillpacks@0.1.1`.

Confirm the exact publish boundary: `skillpacks@0.1.0`, public access, MIT metadata, npm links, and `packages/skillpacks/build` as the publish root.

Confirm the proposed defaults with the user before implementation.

Confirm the public registry does not already contain `skillpacks@0.1.1`.

Convert `alignment/animation-design-planner-deck-builder-transitions.html` to confirmed with read-only decision records and the approval YAML preserved.

Convert the full working packet from raw Markdown preview into readable HTML sections and tables.

Copy each target page to `docs/history/archive/2026-06-07/180623/alignment/` before editing.

Cover prerequisites, first-use flow, generated files, remove/update flow, versioning, published-package verification, and troubleshooting.

Create `alignment/investigate-youtube-owner-analytics-platform.html` as a review-state document-tier alignment page with rendered sections, evidence tables, gates, section feedback, compile YAML, TTS, and question navigation.

Create `alignment/skillpacks-npm-package-walkthrough.html` with confirmed status, document-tier metadata, responsive viewport, TTS include, and current `skillpacks@0.1.0` / `latest` package status.

Create `briefing-slides/create-briefing-slides.html` with slide navigation, review controls, reference links, and YAML compilation.

Create a reusable script under `packages/skillpacks/scripts/`.

Create mirrored base `github-issue`, `github-branch`, and `github-pr` skills with deterministic, low-freedom safety rules.

Create mirrored Codex and Claude `youtube-video-prelaunch-audit` skill directories in `packs/youtube-ops/`.

Create or verify the private GitHub repo `GeorgeQLe/youtube-owner-analytics`.

Create the `prompts/ship/` capture for `$ship`.

Create the local repo layout. **Reconciliation:** single-checkout layout chosen → no local creation/move; three independent remotes exist with `origin` configured on GitHub.

Dangling managed links are reported as missing/updatable and are replaced or removed by the canonical lifecycle command.

Decide and implement the focused validation rule that prevents install-route text from omitting `npx skillpacks install` or `npx skillpacks install-deck` unless explicitly allowlisted.

Define a Codex-only risk-based Sol/Luna/Terra lifecycle without changing Claude skill sources.

Define acceptable treatments such as inline badges, highlighted blocks, or side-by-side before/after sections.

Define compiled responses as a mixed YAML payload containing answered gates, selected `section_feedback`, and unanswered required gate count/status for partial responses.

Define count terms from generated data: platform entries, unique mirrored skills, unique pack skills, unique global skills, active packs, and display cards.

Define issue deduplication, branch naming/adoption, dirty-tree ownership, push, PR upsert, merge, issue-linking, and post-merge cleanup contracts.

Define persona, goal, success condition, and triggering context.

Define runner-specific dual-route wording for Claude, Codex, individual skill installs, pack installs, source-checkout maintenance, and deck installs.

Define the local-first architecture, CLI surface, normalized file layout, evidence contract, scheduling options, and security model.

Define the slide deck workflow: source artifact discovery, dense reference preservation, slide structure, navigation, feedback, gate-answer, annotation, copy, and browser-open behavior.

Delete only `alignment/investigate-youtube-owner-analytics-platform.html` from `agentic-skills`, remove its `alignment/index.html` card, and decrement page/category counts.

Design-tree docs frame Platform Fit as a `user-flow-map` trunk concern and platform probes as thin risk tests, not full parallel products.

Devtool AFPS pack docs exist and identify ORD as a graduation source.

Docs describe cleanup as the path for deprecated artifacts and reserve experimental features for canary/pre-release packages.

Doctor, prune, and remove recognize dangling managed links while preserving unrelated dangling links.

Document and verify the source fix, then commit and push it to `master`.

Document investigation results in `tasks/todo.md`, commit, and push intended changes if tracked files were intentionally modified.

Document results in `tasks/todo.md`, append a brief `tasks/history.md` record, create a ship manifest, and audit task docs.

Document results, commit, and push intended changes while preserving unrelated local edits.

Document results, commit, and push on `master`.

Document results, commit, and push on the primary branch.

Document results, commit, and push the fix on `master`.

Document results, commit, and push the manifest fix on `master`.

Document results, create a ship manifest, commit, and push on the primary branch.

Document results, create a ship manifest, commit, and push the intended changes on `master`.

Document results, produce a ship manifest, commit, and push intended changes on `master`.

Document review/results, produce a ship manifest, commit, and push intended changes on the primary branch if safe.

Document that real `npm publish` is out of scope for this pass.

Document the triage result, commit, and push the fix on `master`.

Dogfood the new path by publishing this migration on its issue-backed non-primary branch as a ready PR; do not merge it automatically.

Draft Flow Assumptions Checkpoint from evidence.

Draft the UI Assumptions Manifest, content requirements manifest, page/entity/action/state matrices, coverage checkpoint, and interview record.

Dry-run and publish `skillpacks` plus `@glexcorp/gskp` as `0.1.22-experimental.6` without moving `latest`; commit, tag, and push release metadata.

Dry-run mode reports intended global and BIP cleanup without changing files.

Each repo is populated and pushed on `master`.

Ensure focused regression coverage catches active chunked skills that combine `Exact next command:` with `## Invoke With YAML`.

Ensure same-version and same-hash installs whose only drift is marker `source=` rewrite the marker without deleting, copying, or logging installed/updated skill lines.

Ensure the skill outputs `specs/user-flow-[topic].md`, `user-flow-[topic]-interview.md`, and `alignment/user-flow-map-{topic}.html`.

Enumerate active `SKILL.md` files under `global/` and `packs/`, excluding `archive/**`.

Enumerate Markdown, HTML alignment, package docs, skill docs, prompt/task docs, and generated docs that should be checked.

Every tracked-file-writing skill inherits or explicitly applies the GitHub delivery contract; no active mutation path pushes directly to `main` or `master`.

Existing valid pinned links and a second idempotent refresh remain unchanged.

Extend `packages/skillpacks/src/cli/lifecycle.mjs` cleanup to remove managed project-local deprecated alias installs and preserve unmanaged directories.

Extend `prepublish-auth-check.mjs` with package/version overrides and non-interactive web-auth rejection when no publish token or legacy/token auth is detected.

Extend `publish.sh` with `--tag <dist-tag>` and `--preid <identifier>` options, keeping `latest` as the default tag for stable releases.

Extend `scripts/audit-alignment-pages.mjs` with BIP gate-sequencing diagnostics for premature final gates, stale future-channel drafting wording, and stale all-channels-not-now selected-channel drafting options.

Extend `tests/layer1/post-uat-consolidation-routing.test.ts` for the new verification contract.

Extend active alignment-page audit behavior to read `.agents/project.json`, require BIP handling for active Stage 2 review pages when BIP is enabled, avoid false positives for Stage 1/confirmed pages, and include active BIP pages in `alignment/index.html`.

Extend active-page audit behavior and layer1 fixtures/tests to require a hidden agent answer per open question.

Extend convention bundle audit canaries to fail when briefing-first dense-backup language disappears.

Extend focused layer-1 routing coverage.

Extend focused layer1 tests for schema, sample, route stability, platform-probe validation, and mirror contracts.

Extend planning, execution, independent review, quality gating, and shipping contracts.

Extract the approved proposed spec and interview log from `alignment/user-flow-map-skill-execution-handoff.html`.

Final status shows only intended changes plus any pre-existing unrelated files.

Fix MIT license metadata, npm repository links, staged package metadata, and package metadata coverage.

Fix package verification blockers found during readiness checks.

Fleet dry-run reviewed 233 installs, 389 updates, 38 managed removals, and three preserved unmanaged links before mutation.

Fleet refresh reported `60 ok, 0 flagged, 0 failed`; `doctor --all` reported the same summary.

Flow coverage was reviewed in the prior conversation context.

Focused affected test(s), if a patch is made.

Focused lifecycle suite passed 76/76 and the full package Node suite passed 217/217.

Focused product-design test coverage enforces the new behavior.

Focused tests cover the new path and required context fields.

For confirmed true research producers, keep the 4-step staged workflow and correct lifecycle metadata.

For misclassified skills, change `type:` to the correct existing category while preserving intended behavior.

Forward current `pack.sh` commands while preserving consumer project `cwd`.

Framework mirrors keep parent-owned routing and output paths unchanged.

Generate `packages/skillpacks/dist/skillpacks-manifest.json` from `global/`, `packs/`, `PACK.md`, and `SKILL.md` frontmatter.

Generate `research/researchish-skill-lifecycle-audit.md` from audit output.

Generated alignment convention text names `alignment/bip/{skill-name}.html`, including archive paths under `docs/history/archive/YYYY-MM-DD/HHMMSS/alignment/bip/{skill-name}.html`.

Generated catalog exports validate.

Generated fixture pages pass the existing active-page auditors.

Generated package/catalog artifacts reflect the active-surface pruning.

Generator checks must stay clean after future upgrades.

Group targets by pack/domain for staged future execution.

Harden `consolidate-prototypes` against not-ready UAT files, unchecked readiness items, all-`Not run`/deferred evidence, and unbuilt approved branches without explicit user handling.

Harden `consolidate-prototypes` evidence gates for not-ready UAT files, unchecked readiness items, all-`Not run` logs, and unbuilt approved branches.

Help text, compatibility docs, and tests document `cleanup --global` as the replacement for device-wide `uninstall-global` usage.

Identify any active skill that still requires `pack` to install packs or skills.

Identify canonical sources versus generated or historical artifacts to avoid false positives.

Identify install-routing, pack-availability guard, and missing-skill fallback references.

Identify minimal high-confidence cleanup actions and larger follow-up remediation work.

Identify non-blocking legacy/fallback references that should be rewritten before archival.

Identify that the tracked CLI entrypoint imports an untracked `update-check.mjs` module.

Identify the minimal fix and add focused regression coverage.

Identify the root cause and affected layout rules.

Identify the root cause with file/line evidence and choose the minimal fix.

If no low-risk patch is accepted, the measured bottleneck and rejected rationale are documented without source changes.

Implement `cleanup` as the primary command with `uninstall-global` preserved as a compatibility alias.

Implement `init-global` by invoking packaged `init.sh`.

Implement `lstat`-based dangling-link inspection and current/legacy managed-link ownership.

Implement Phase 1 safety subskills and the canonical GitHub delivery contract.

Implement Phase 2 shipping-orchestrator migration over the validated `github-issue`, `github-branch`, and `github-pr` contracts.

Implement Phase 2 shipping-orchestrator migration.

Implement Phase 3 provisioning, docs, audits, generated artifacts, and verification.

Implement the smallest durable CLI/test change for legacy repo-managed global install cleanup or reporting.

In a fresh temp project, install one pack, one individual skill, and one deck from the published package.

In the linked-BIP branch, keep the existing reference and handoff checks and fail when linked BIP coexists with an active final artifact approval gate.

Include `packages/skillpacks/dist/skillpacks-manifest.json` in the package allowlist.

Include `SKILL.md`, `CHANGELOG.md`, `agents/openai.yaml`, and generated `ALIGNMENT-PAGE.md` for both mirrors.

Include approval gates for evidence coverage, assumptions, flow map, branch/state/handoff coverage, artifact destination, proposed file changes, and downstream route.

Include counts by category.

Include deck metadata for `vard`, `ord`, `business-afps`, and `devtool-afps` with package-list and registry-tag fields.

Include pack names, skill names, tools, versions, content hashes, archive versions, source paths, and status.

Initialize `/Users/georgele/projects/tools/youtube-owner-analytics` as an independent git repo on `master` with `origin=https://github.com/GeorgeQLe/youtube-owner-analytics.git`.

Inspect `publish.sh`, existing publish recovery tests, package metadata, and current dirty release files without overwriting unrelated work.

Inspect `publish.sh`, focused publish recovery tests, package metadata build behavior, and recent commits around dirty-tree handling.

Inspect `publish.sh`, the npm auth preflight script, the npm debug log, and the existing publish recovery tests.

Inspect `scripts/skill-links.sh`, focused shell installer tests, and package lifecycle coverage.

Inspect `uninstall-global`, BIP config storage, docs, and tests.

Inspect active source surfaces: pack metadata, active skills, docs, routing maps, catalog/showcase generation scripts, and generated exports.

Inspect BIP language in `CLAUDE.md`, generated alignment convention bundles, `ship`/`ship-end` skills, tests, and repo status.

Inspect both Codex and Claude `investigate` mirrors, changelog, and archive requirements.

Inspect current `build-ui-screens` mirrors, generated copies, task docs, and dirty worktree state.

Inspect current `idea-scope-brief` mirrors, deck docs, changelogs, and routing tests.

Inspect current `packages/skillpacks/src/cli/lifecycle.mjs`, package scripts, and lifecycle tests around refresh, dry-run, install updates, and reload notices.

Inspect current `skillpacks` CLI, pack lifecycle, init, docs, and git history.

Inspect current BIP evidence, canonical alignment convention, generated `ALIGNMENT-PAGE.md` bundles, audit scripts, package command surface, and representative alignment-producing skills.

Inspect current briefing, alignment, and interrogation convention wording plus generator/audit surfaces.

Inspect current Codex/Claude `youtube-video-prelaunch-audit` skill files, changelogs, archives, and relevant audit scripts.

Inspect current generator, audit, package, and test surfaces.

Inspect current interrogation convention, generated `INTERROGATION-PAGE.md` bundles, audit script, layer1 tests, and upgrade skill mirrors.

Inspect current provision skill variants, root shipping contract wording, lifecycle tests, archives, and generated package scripts.

Inspect current public docs and `publish-canary` helper output for ambiguous experimental-channel examples.

Inspect current publish, auth-check, and recovery test behavior.

Inspect current rapid deck docs, VARD/ORD skill mirrors, pack metadata, generated references, and task notes.

Inspect current repo state, package version, changelog, and post-`0.1.18` commits.

Inspect current schema, sample manifest, design-tree convention, mirrored skills, tests, and active task docs.

Inspect current skill mirrors, pack metadata, conventions, generated bundles, references, and tests.

Inspect current tree, docs, git status, and split commit history.

Inspect existing `uninstall-global`, BIP config helpers, docs, and focused tests.

Inspect existing CLI namespace, package staging, and package boundary tests.

Inspect existing dirty work and routing handoff hits without reverting unrelated changes.

Inspect existing dirty work and routing handoff hits.

Inspect existing pack, deck, install, and versioning docs.

Inspect git status, unpushed commits, task docs, manual tasks, advisory tasks, and deploy contract.

Inspect install resolution code, tests, and recent history.

Inspect lifecycle CLI, provision templates, routing audit, docs, and existing task state.

Inspect package version state, tags, npm registry state for both package names, release scripts, deploy contract, and post-`v0.1.13` commit history.

Inspect Pattern A final handoff and research-health routing surfaces for places that currently default to full reruns after small post-canonical corrections.

Inspect relevant research, existing user-flow specs, UX decisions, app routes, and handoff/operating-mode docs.

Inspect repo guidance, lessons, current task state, and active `SKILL.md` inventory.

Inspect repository instructions, lifecycle ownership paths, regression coverage, current release state, and fleet-recovery constraints.

Inspect the actual global and project-local `codebase-status` skill files and install markers.

Inspect the approved npm distribution design and existing script contracts.

Inspect the canonical alignment-page convention, generated `ALIGNMENT-PAGE.md` bundles, audit script, layer1 tests, recent history, and active/archived Alignmeant BIP page state.

Inspect the canonical BIP convention, social-output routers, audit script, and focused tests.

Inspect the current `ui-interview` review page and working packet.

Inspect the current BIP alignment convention, social post/video conventions, `idea-scope-brief` Codex and Claude skill instructions, audit diagnostics, and package tests.

Inspect the current BIP checkpoint audit, alignment-page convention, generated bundle behavior, and layer1 fixtures.

Inspect the existing `gskp alignment` CLI namespace, package staging boundary, and tests.

Inspect the existing `publish-skills` wrapper and release runbook canary flow.

Inspect the existing alignment page, session insight memory, and available Claude/Codex history sources.

Inspect the failing package-boundary and publish-recovery tests.

Inspect the Graduation Path chart markup and CSS.

Inspect the slowest repeat-work path and patch only confirmed duplicate or avoidable work while keeping at least one full staging and one full `npm pack --dry-run` for each distinct lane/package-state combination under test.

Inspect the staged tarball boundary and denied paths.

Inspect the three target alignment pages and `compile-central-alignment` contract.

Interrogation generator/test registry includes all five skills and expects 46 participating stubs.

Inventory active Codex and Claude `SKILL.md` files and define mismatch patterns.

Inventory entry points, preconditions, happy path, alternate paths, branch rules, screens/routes, failures, recovery, and handoffs.

Inventory the existing direct-primary shared contract, shipping orchestrators, `commit-and-push-by-feature`, and legacy `branch-lifecycle` behavior.

Keep `init-global` as the explicit user-home global install path.

Keep `npx skillpacks doctor` read-only.

Keep `pack.sh` and Node command output aligned for supported aliases.

Keep `research/.progress.yaml` scoped to product-path/product-line state and out of UAT/prototype/consolidation readiness storage.

Keep `scripts/pack.sh` as a wrapper or tested fallback.

Keep `scripts/skill-pack-routing-audit.sh` focused on cross-pack availability guards unless extending it is cleaner than adding a dedicated npm-route check.

Keep `uninstall-global` as a compatibility alias while routing docs toward `cleanup --global`.

Keep BIP commands deprecated/failing and kanban packs/skills non-installable.

Keep canonical UI requirements files unwritten.

Keep COA A as the first package transport while making `install-deck` a resolver over COA B/C-compatible metadata.

Keep command ownership/dependency docs aligned with the CLI compatibility matrix.

Keep downstream research routing as secondary context after deck selection.

Keep existing `pnpm --dir apps/skills-showcase ...` commands working during the transition.

Keep final approval limited to `approval_status: ready-for-agent-review` only when all required gates are answered and no unresolved negative or clarification feedback remains.

Keep gate questions inline and preserve filmstrip gate markers.

Keep generated outputs under `apps/skills-showcase/public/assets/`, the temporary `docs/skills-showcase/assets/` mirror, and the website-owned benchmark matrix.

Keep index counts and metadata consistent with the existing card format.

Keep staged package manifests tied to the selected release lane instead of stale `dist` state.

Keep the audit script read-only except when report generation is explicitly redirected by the agent.

Keep the catalog layer read-only.

Leave `alignment/skills-inventory.html` untouched as a generated/static catalog.

Leave P1/P2 remediation files unchanged when no drift is found.

Leave remove behavior unchanged.

Legacy absolute links from moved `agentic-skills`/`skillpacks` checkouts are recognized only when tool and skill layout match.

Link to the current Markdown docs for deeper reference.

List every `misclassified` skill.

List every alignment-page skill that appears to belong in `scripts/alignment-skip-list.txt`.

List every marker-compliant `type: research` skill that is semantically suspicious.

List every non-research skill with `research/` output language.

List tracked Markdown/HTML/docs artifacts across root docs, research, specs, alignment, tasks, benchmark, and pack/global skill docs.

Load the `pack` skill instructions because the request names the skill.

Locate canonical design-tree-loop source and active state-model mirrors.

Locate the existing `youtube-video-audit` skill and confirm its post-release performance-audit scope.

Make `doctor` and `prune` account for project-local base skills without deleting unmanaged user content.

Make clear that change indicators must be visible in the HTML itself.

Make high-confidence deck fit the primary `## Next Steps` recommendation, with canonical deck installs using `npx skillpacks install-deck <deck>`.

Make publish-recovery tests valid when the source package version is already a prerelease during the release dry run.

Make source-checkout CLI execution and staged package execution both resolve `pack.sh` and `init.sh` correctly.

Make the fallback order explicit: helper script when present, WSL PowerShell `file://wsl.localhost/<distro>/...` bridge when available, `xdg-open` when available, then `failed` with absolute path.

Make the source version bump transactional until the first real `npm publish` starts.

Mark completed current-task headings as historical/no-active state.

Mark docs as current, needs update, duplicate/superseded, generated/local-only, or archive candidate.

Mark the old npm distribution strategy page and index card as historical/superseded for package usage, with the current walkthrough as the usage reference.

Mirror the carve-out in the codex skill — `packs/product-design/codex/ux-variations/SKILL.md` (§0c Required Progress Handoff Block lines ~82,89–92; step 8 line ~285) — assemble-ready leads with the HTML review path + single compiled response YAML; no same-command rerun line, no second YAML.

Mirror the same carve-out in the claude variant — `packs/product-design/claude/ux-variations/SKILL.md` — identical change with `/ux-variations` syntax.

Mirrored skills mention Platform Fit Workshop, broad candidate set, platform probes, and `spec-interview` final production platform lock where applicable.

Move `base/` immediately while preserving legacy managed-marker cleanup for old `base/...` source paths.

Move real-publish package/auth preflight ahead of source mutation by computing the target version in a temporary package copy.

Move Skills Showcase data generators and validator under `apps/skills-showcase/scripts/`.

Move slide feedback, mark, and annotation controls into the active-slide sidebar.

Move the `skillpacks` CLI and source code under `packages/skillpacks/`.

Move the deck-based installation section before the review gates so it is part of the reviewed body.

Move tracked base skill sources and archives to `packs/base/{claude,codex}` and add `packs/base/PACK.md`.

No active updated handoff recommends installing the `uat` skill directly.

No application repository is committed or pushed, and unrelated dirty files are byte-for-byte unchanged.

No clear duplicated paragraphs remain within an edited active `SKILL.md`.

No edited active `SKILL.md` repeats substantial prose from its direct Markdown references when a pointer or shorter instruction is sufficient.

No nested git repos are accidentally tracked by another repo (no local sibling clones created).

No regressions in previous phase tests.

No stable `v0.1.22` tag is created and npm `latest` is not moved.

No unrelated files are removed from the current checkout.

Normalize repo roots. **Reconciliation:** verified remotely (showcase app at root; benchmarks harness at root).

npm dist-tag and published smoke verification for both package names

Open or attempt to open the alignment page in the browser.

ORD graduation routes to `npx skillpacks install devtool`, then `$devtool-workflow` by default or `$devtool-user-map` as the first concrete research step.

Package and manifest metadata check for `0.1.22-experimental.1` / `canary`

Package asset content checks for briefing/alignment/interrogation convention assets.

Package manifest and skills-catalog export artifacts are regenerated from current source.

Package manifest output reflects the bumped provision skill metadata.

Package the templates through build staging and npm `files`.

Patch `sync_skill_install` copy behavior to exclude `archive/` directories at every depth while preserving pinned archive symlinks.

Patch active `SKILL.md` report-first and staged-research wording to require scope approval before synthesized research.

Patch clear platform-convention mismatches with minimal edits.

Patch convention and mirrored skill handoff wording.

Patch only clear same-document or skill/reference duplication with minimal wording changes.

Patch release-blocking findings only; archive/version `SKILL.md` behavior changes and regenerate package/catalog metadata after patches.

Patch shared convention docs so dense HTML pages are produced first, decks are produced second, only decks are opened when the briefing convention exists, and compiled YAML routes back to the producing skill.

Patch stale canonical or installed mirrors that still require `Exact next command` alongside YAML.

Patch stale canonical or installed skill mirrors that still require an `Exact next command` alongside YAML.

Patch the `ship-end` BIP step so enabled BIP skips only the enablement prompt, then prompts the agent to draft supported post suggestions after the wrap-up report.

Patch the canonical alignment-page convention and focused layer-1 tests.

Pilot dry-run, refresh, and doctor passed; its 32 dangling managed links fell to zero and unrelated dirty-file hashes were unchanged.

Port `.agents/.pack.lock` handling.

Port `doctor`, `pin`, `unpin`, and `prune`.

Port pack alias normalization and hibernated-pack diagnostics.

Port pack and individual skill install/remove/refresh logic.

Preserve `--current` partial-publish recovery while reusing the intended dist-tag and publishing only the missing package.

Preserve `.agentic-skills-managed` marker creation and content hash behavior.

Preserve `pack.sh` forwarding for all existing commands.

Preserve `scripts/pack.sh` where the text is explicitly about source-checkout maintenance.

Preserve `skillpacks init` project-local install behavior.

Preserve approved decisions: `skillpacks`, hybrid COA A first, skill-level pinning, deck presets first.

Preserve Claude `/exec`'s dirty-tree handoff while requiring `/ship` to create or adopt the issue-backed branch before committing.

Preserve existing entry dates and define how to fill missing dates.

Preserve existing partial-publish behavior when `skillpacks@$VERSION` exists and `@glexcorp/gskp@$VERSION` is missing.

Preserve existing status reporting semantics: `focused`, `opened`, `fallback-opened`, `blocked`, or `failed`.

Preserve explicit "no real publish in Phase 4" language.

Preserve hibernated pack/skill safety diagnostics before unknown-name failure.

Preserve historical counts only when explicitly labeled historical/prototype scope.

Preserve intentional compatibility behavior for `business-app`, `creator-media`, and `business-discovery` aliases.

Preserve layout/styling and implementation as non-goals.

Preserve lightweight VARD/ORD scan criteria while moving candidate ranking into approved preliminary packets.

Preserve partial response behavior when at least one gate answer or one section feedback item exists.

Preserve post-write `## Recommended Next Command` routing after approved artifacts are written.

Preserve rejection for unsupported `init` arguments such as `--bad`.

Preserve runner syntax exactly: Claude gets `/pack ...`; Codex gets `$pack ...`; shell guidance gets `npx skillpacks ...`.

Preserve safety boundaries: no draft posts, video ideas, or channel-specific content before channel-selection approval; each recommendation needs source basis, fit rationale, claim risk, and non-recommended-channel rationale.

Preserve strict default behavior and preserve the narrow `--current` release-metadata recovery exception.

Preserve the no-context-loss requirement without using one raw `<pre><code>` Markdown dump as the primary review surface.

Preserve the non-research `research/` and alignment skip-list candidate queues as report inventory for a later review batch instead of applying broad heuristic edits.

Preserve unrelated project config fields on writes.

Preview bundle changes with `node scripts/upgrade-alignment-page.mjs --dry-run`.

Print unified diffs before writes and create timestamped `.agents/backups/` copies before agent-doc mutation.

Progress handoff session guidance says to paste the YAML only into the fresh/clean context alongside the exact command.

Promote the high-confidence target list into `tasks/todo.md` as the active task.

Promote this implementation plan into `tasks/roadmap.md` and `tasks/todo.md`.

Publish + manifest (done in `/ship`) — staged source first, regenerated the public catalog export (`generate-skills-catalog-export.mjs` + validate → fresh) and the index-generated `skillpacks-manifest.json` (`build:manifest`; `build:check` passed: 411 skills, 383 bundles), committed source + bundles + catalog + manifest atomically, then ran `scripts/pack.sh refresh` to republish the untracked local `.codex/skills/` `.claude/skills/` roots (gitignored, not committed).

Publish recovery tests pass from both stable source versions and temporarily pre-bumped experimental source versions.

Publish the completed migration as a ready PR without merging it automatically.

Published `.6` fleet refresh dry-run, refresh, and targeted doctor/dangling-link audit

Published `.6` pilot refresh dry-run, refresh, and doctor

Published `skillpacks@experimental` deck-backed `install-deck game-afps` -> `remove game` -> `doctor`

Rank deck candidates by domain, tempo, and concept signals.

Re-check npm registry state for `skillpacks`.

Re-check safe npm registry/account information for `skillpacks` without publishing.

Re-read the shared alignment-page convention, existing tests, and active affected skill list.

Re-run dry-run and confirm no pending updates.

Re-run scoped stale-count and retired-route scans.

Re-run scoped stale-route, install-wording, publication-wording, historical-label, wrapper, alignment-page, generated-bundle, focused layer1, and whitespace checks.

Re-run skill-map generator syntax and regeneration.

Re-run the target derivation search and confirm this todo list matches the high-confidence set.

Read `$exec`, `animation-design-planner` approval handling, task docs, app docs, and the approved alignment page.

Read `$targeted-skill-builder`, active lessons, current `idea-scope-brief`, deck docs, repo config, and task docs.

Read `prompts/pack/skill-prompt-20260610-195858-skillpack-routing-audit.md`.

Read active `ui-interview` instructions and bundled alignment contract.

Read active lessons and relevant task docs.

Read active lessons and task docs.

Read active repo instructions, lessons, task docs, and relevant alignment-page/test context.

Read current init/init-global routing, docs/help, tests, and package scripts.

Read product-design skill conventions, routing docs, research-roadmap rules, alignment generator, and benchmark coverage patterns.

Read project lessons, the alignment-page convention, source research brief, existing alignment examples, and current git status.

Read relevant lessons and task state.

Read source evidence from the confirmed flow spec, Skills Showcase idea brief, deck-creation flow, operating-mode/approval-packet docs, and relevant app route/component files.

Read the `investigate` skill, current git state, source research brief, source alignment page, alignment index, and prior prompt history for the exact next-step YAML source.

Read the active `$targeted-skill-builder` and `$session-triage` contracts.

Read the active `pack` skill instructions.

Read the active `ui-interview` skill instructions and bundled alignment-page contract.

Read the active `user-flow-map` skill instructions.

Read the approved alignment page and gate answers.

Read the four active scan skills and the reference staged research pattern.

Rebuild `packages/skillpacks/build`.

Rebuild the staged package output after source edits.

Receive confirmation or corrections.

Recheck the deck-backed remove/doctor path that failed in `0.1.22-experimental.4`.

Recommend whether archiving/sunsetting is safe now or after specific follow-up edits.

Reconcile `tasks/todo.md` so task-doc audit sees completed review evidence rather than stale unchecked verification items.

Reconcile stale task bookkeeping for the already-shipped install-destination output correction.

Reconciled as historical follow-up, not active executable work in this task file.

Record archive paths and replacement rationale in task history.

Record closeout evidence and ship the source state.

Record closeout evidence in task history and a ship manifest.

Record closeout evidence in task/history docs.

Record current handoff checklist in task docs.

Record current-phase work and review notes in `tasks/todo.md`.

Record current-phase work in `tasks/todo.md`.

Record dry-run output in the ship manifest.

Record history and the quality-gate manifest.

Record history, ship manifest, per-repository before/after path changes, final verification, and review.

Record implementation progress and validation results in `tasks/todo.md`.

Record measurements, patch rationale, final verification, and ship intended changes.

Record release evidence in a ship manifest.

Record review and ship intended changes without reverting unrelated dirty work.

Record review evidence and ship intended tracked changes.

Record review notes and ship intended changes.

Record review notes and ship only intended review-page changes.

Record review notes in `tasks/todo.md`.

Record review notes, commit, and push intended changes on the primary branch unless blocked.

Record review notes, commit, and push intended changes.

Record review notes, history, and a quality-gate ship manifest.

Record review notes, history, and quality-gate manifest.

Record review notes, history, and ship manifest.

Record review notes, then commit and push only intended changes.

Record review notes, update history, commit, and push intended changes only.

Record review notes, update history, commit, and push intended remediation artifacts only.

Record review results in task docs, then commit and push intended changes.

Record review, commit, and push intended changes.

Record review, commit, and push.

Record review, ship manifest, and history for this base migration before starting `research-amend`.

Record scope boundaries so artifact-only skills, package metadata, archives, changelogs, and skill source refactors stay out of this tracking-only step.

Record starting `git status --short --branch --untracked-files=normal` and preserve the completed canary closeout source/task state already in the tree.

Record task plan and prompt history for the UAT pack availability guard work.

Record the active implementation phase in `tasks/todo.md`.

Record the audit plan in `tasks/roadmap.md` and `tasks/todo.md`.

Record the clean rerun in task docs, history, and ship manifest.

Record the full phased migration and acceptance criteria in `tasks/roadmap.md`.

Record the review, commit, and push intended tracked changes.

Record the revision plan in `tasks/roadmap.md` and `tasks/todo.md`.

Record the user correction in `tasks/lessons.md`.

Record the user correction in lessons and prompt history.

Record the user correction in lessons and task state.

Record this active roadmap/todo plan before implementation.

Record this plan and the active todo before implementation.

Record this roadmap and active todo before implementation.

Record this roadmap/todo plan before implementation.

Record this validation plan before repo-wide scanning.

Record validation and review results in `tasks/todo.md`.

Record validation results in `tasks/todo.md`.

Record verification in `tasks/todo.md`, then commit and push intended changes.

Record verification results and final repo status.

Record verification results, review diffs, commit, and push intended changes.

Recreate the canary source state for `packages/skillpacks` at `0.1.22-experimental.1`.

Refactor `branch-lifecycle` into a compatibility/advanced-recovery wrapper over `github-branch` and `github-pr` rather than an exception to direct-to-primary development.

Refactor `commit-and-push-by-feature` into a compatibility wrapper over safe non-primary branch publication.

Refactor alignment and interrogation generators to write/validate stubs and resolver metadata instead of generated sibling bundles.

Refactor mirrored `ship` and `ship-end`, plus Codex `exec`, so task docs, commits, branch push, and PR readiness happen in one review boundary.

Refresh `/Users/georgele/projects/web/dev/vectorfit-redux` with the fixed source-checkout installer and verify no nested archives remain.

Refresh fixtures to prove valid/invalid install route behavior.

Refresh generated bundles and run targeted verification.

Refresh generated package artifacts.

Refresh generated package/catalog metadata when required.

Refresh generated showcase data if required by skill metadata changes.

Refresh generated/runtime copies if needed, validate the skill, and run focused repository checks.

Refresh generated/runtime skill copies, including the active pair-app runtime copies through the supported refresh path.

Refresh page/index metadata only where needed.

Refresh replaces dangling Claude and Codex managed links with managed directories, including legacy `/home/.../agentic-skills/packs/...` links.

Refresh skills catalog exports required by skill changes while preserving unrelated package metadata.

Refresh Skills Showcase data if required by changed skill metadata.

Refresh Skills Showcase generated data if active `SKILL.md` behavior/metadata changes require it.

Refresh stale README wording that said the canary lane did not exist yet.

Refresh the managed `gblock-party-redux` installed `ui-interview` copy from the updated source pack.

Refresh the managed `gblock-party-redux` installed skill copy from source.

Refresh the runtime skill copies and generated skill/catalog manifests without committing generated `.claude/skills/**` or `.codex/skills/**` roots.

Refuse missing, duplicate, malformed, or unknown provision markers safely.

Regenerate `exports/skills-catalog/v1/` public export artifacts.

Regenerate `packages/skillpacks/dist/skillpacks-manifest.json` from the current git index.

Regenerate `packages/skillpacks/dist/skillpacks-manifest.json` with `SKILLPACKS_PACKAGE_LANE=canary`.

Regenerate `packages/skillpacks/dist/skillpacks-manifest.json`.

Regenerate alignment convention bundles and packaged assets as needed.

Regenerate alignment-page bundles with `--legacy-bundles` and verify generated resolver stubs remain in sync.

Regenerate and validate Skills Showcase data.

Regenerate bundled `ALIGNMENT-PAGE.md` files from the canonical convention.

Regenerate bundled `ALIGNMENT-PAGE.md` files from the canonical source.

Regenerate bundled `ALIGNMENT-PAGE.md` files.

Regenerate bundled alignment-page conventions from `docs/alignment-page-convention.md`.

Regenerate bundles — ran `node scripts/upgrade-design-tree-loop.mjs` (22 `DESIGN-TREE-LOOP.md` bundles rewritten with the assemble-ready exception) and `node scripts/upgrade-alignment-page.mjs`. Note: in default mode the alignment script updates SKILL.md resolver stubs only and does **not** rewrite the legacy sibling `ALIGNMENT-PAGE.md` bundles (those are frozen fallback artifacts, not `--check`-validated by default); the alignment convention edit propagates to runtime via the packaged convention resolver, regenerated at build/publish time (`/ship`). Both `--check` gates exit 0.

Regenerate design-tree-loop bundles.

Regenerate generated `ALIGNMENT-PAGE.md` bundle files with `node scripts/upgrade-alignment-page.mjs`.

Regenerate generated `ALIGNMENT-PAGE.md` bundles and run the requested focused verification.

Regenerate generated `ALIGNMENT-PAGE.md` bundles from the canonical convention.

Regenerate generated alignment-page artifacts.

Regenerate generated interrogation-page bundles from the canonical convention.

Regenerate legacy `INTERROGATION-PAGE.md` bundles from the canonical convention.

Regenerate legacy `INTERROGATION-PAGE.md` bundles from the canonical source with the generator.

Regenerate or verify `alignment/index.html` locally and keep it untracked unless repo convention forces otherwise.

Regenerate or verify packaged/runtime convention copies as needed without adding `briefing-slides` to every producer skill.

Regenerate package build docs/manifest if `build:check` requires it.

Regenerate package manifest/catalog exports from staged source and run focused verification.

Regenerate package manifest/catalog exports if source discovery changes.

Regenerate package staging through the skillpacks build.

Regenerate package/catalog artifacts and review diffs.

Regenerate package/catalog artifacts if any tracked `SKILL.md` or `PACK.md` metadata/content changes.

Regenerate package/catalog metadata.

Regenerate packaged convention assets through the package build path.

Regenerate public skills catalog export metadata.

Regenerate release metadata:

Regenerate showcase data and validate the public skill catalog artifacts.

Regenerate skill-map artifacts after moving `game` into its own Game AFPS row.

Regenerate/check bundled `DESIGN-TREE-LOOP.md` files.

Regenerate/check bundled design-tree convention files.

Register and package the `briefing-slides` convention.

Register the parent and framework subskills in the interrogation generator/test contract.

Reject `cleanup --all --global` with a clear error.

Release and deploy skills require an already-merged, current primary branch and never reinterpret an open PR as deployable primary state.

Remediate the active `gblock-party-redux` incident review page found during spot-check.

Remediate the active `gblock-party-redux` review page still carrying the incident pattern.

Remove active `prototype`, `create-ui-experiment`, and `consolidate-variations` source directories from `packs/product-design`.

Remove active discovery, recommendation, or default-route references for cut surfaces.

Remove BIP project config keys during cleanup across discovered projects, with dry-run reporting and no unrelated config loss.

Remove direct or compressed `logic-wiring` route from prototype approval to consolidation; route only to UAT evidence capture.

Remove direct-primary overrides from active writing skills and point them at the shared GitHub delivery contract.

Remove generated untracked `.agents/skillpacks/` runtime docs from the release boundary.

Render the full proposed interview log inline in the review alignment page.

Render the full proposed spec inline in the review alignment page.

Render the results into the requested alignment HTML page, including counting rules, evidence samples, charts/tables, and recommendations.

Replace `experimental-skill` placeholder examples with `<pack-or-skill>` and add stable/default install examples.

Replace retired executable `icp` route examples in current docs, specs, and indexed alignment pages with `customer-discovery`, while preserving `enterprise-icp` and `research/icp.md` artifact references.

Replace the separate bottom `Compile Feedback YAML` and `Compile Answers` model with one bottom `Compile Responses` control.

Representative research and product-design skills inherit the behavior through shared conventions without direct `required_conventions` churn.

Reproduce or confirm the right-edge clipping at relevant viewport widths.

Reproduce or model the dirty-state failure path with the existing test harness or a safe dry-run path.

Require `YYYY-MM-DD` dates in muted meta spans after index links.

Require a `Test And Compare Launch Set` containing exactly three simultaneous title/thumbnail pairs, each with full title, thumbnail concept, packaging hypothesis, intended audience signal, and win implication.

Require confirmed pages to remove the final `Compile Responses` button, response counters, input controls, and approval-blocking UI.

Require local section feedback textarea, local compile/copy controls, and read-only YAML output to be hidden until emphasize, thumbs-down, or clarification is selected for that section.

Require structured HTML rendering of the working packet sections, especially Markdown tables.

Required gate slides have a documented root marker and `data-gate-status` contract.

Required radio options carry explicit `data-approval-effect="approve" | "block" | "clarify" | "other"` semantics.

Rerun final release gates before commit/push and publish dry run.

Rerun package build/check and focused package tests.

Research official YouTube Data API v3, YouTube Analytics API, and YouTube Reporting API fit from primary documentation.

Resolve active product path to `skills-showcase`.

Resolve exact active pack slugs next.

Resolve exact active pack titles using trimmed/collapsed whitespace title matching.

Resolve install tokens by exact active skill name first.

Resolve the published package through `npx --package skillpacks@latest`.

Restore only `packages/skillpacks/package.json` and `packages/skillpacks/dist/skillpacks-manifest.json` from the failed `0.1.15` bump back to `HEAD`.

Restore the current leftover failed-publish `0.1.14` source bump to the committed `0.1.13` state.

Restore the current stranded `0.1.14` package metadata back to `0.1.13`.

Retry only npm propagation-style `npx --package "$NPM_SPEC"` failures such as `ETARGET`, `notarget`, or `No matching version found`; fail immediately for real CLI/test errors after package resolution.

Reuse one full-deck YAML builder for the sidebar all-feedback compiler and slide 8 compiler.

Reuse the existing YouTube research-stage, feedback, alignment, handoff, and pack-availability conventions.

Review artifacts present complex interfaces progressively, with clear first-step and primary-path focus before dense secondary controls.

Review gates cover final artifact approval, MVP scope, Reporting API timing, evidence privacy, and unresolved open questions.

Review generated diff and run whitespace checks.

Review high-signal duplicate findings and classify each as intentional mirror, generated convention copy, or redundant local prose.

Review the diff for unintended generated drift.

Review the diff for unrelated changes and intended scope.

Review the fleet dry-run, repair all discovered projects through the published CLI, verify the expected summary and formerly failing projects, and report per-repository path changes without committing application repositories.

Revise `docs/alignment-page-convention.md` inside the `alignment-convention` markers.

Revise alignment-page auditing and package-boundary expectations for the new post-confirmation BIP page shape and exhaustive channel behavior.

Rewrite `## Next Work` and routing wording so missing prototype build plans route to `user-flow-map --prototype-build-plan`, while direct `logic-wiring` requires a plan item with the approved `ui_experiment_id` or an explicit user-accepted untracked ad hoc bypass.

Rewrite the root `package.json` as private workspace metadata for `agentic-skills`.

Route `list-packs`, `status`, `set-mode`, and `set-update-mode` through Node without requiring `bash` or `jq`.

Route `skillpacks init --global` to packaged `init.sh` through the same global-core path as `init-global`.

Run `git diff --check`.

Run `node scripts/audit-alignment-pages.mjs`, `node scripts/audit-task-docs.mjs`, and `git diff --check`.

Run `node scripts/audit-alignment-pages.mjs`.

Run `node scripts/audit-task-docs.mjs`.

Run `node scripts/upgrade-alignment-page.mjs --check`.

Run `node scripts/upgrade-alignment-page.mjs --dry-run`.

Run `node scripts/upgrade-alignment-page.mjs` to refresh generated bundles.

Run `node scripts/upgrade-alignment-page.mjs`.

Run `node scripts/upgrade-interrogation-page.mjs --check`.

Run `npm --workspace skillpacks run build:check`.

Run `npm --workspace skillpacks run build`.

Run `npm --workspace skillpacks run test:node`.

Run `npm --workspace skillpacks run verify:package`.

Run `npm pack --dry-run`.

Run `npm publish --access public` only after explicit confirmation.

Run `npm publish --dry-run --json` from `packages/skillpacks/build`.

Run `npm publish --dry-run` locally without publishing.

Run `npm whoami` against the npm registry and record whether auth is ready.

Run `pnpm --dir tests exec vitest run --project layer1 layer1/alignment-gates.test.ts`.

Run `pnpm --dir tests exec vitest run --project layer1 layer1/audit-alignment-pages.test.ts`.

Run `pnpm --dir tests exec vitest run --project layer1 layer1/upgrade-alignment-pages.test.ts`.

Run `scripts/skill-archive.sh` for the four scan skill directories.

Run `scripts/skill-install-routing-audit.sh --active`.

Run `scripts/skill-install-routing-audit.sh --fixtures tests/fixtures/skill-install-routing`.

Run a dry-run publish preflight for the next patch version without leaving release-state mutations behind.

Run active-skill audits, archive/version checks, diff checks, and status checks.

Run alignment audit, packaged alignment audit, focused package tests, dry-run packaging, task-doc audit, and diff hygiene.

Run alignment-page audit, root wrapper smoke checks, and whitespace checks.

Run alignment-page audit.

Run benchmark coverage validation and focused layer1 setup tests.

Run convention audits, package build checks, package-content checks, and spot-check inherited skill behavior.

Run convention, page, package freshness, and focused layer1 regression checks.

Run focused after timing and full validation gates.

Run focused and package verification.

Run focused and package-facing verification.

Run focused and package-level verification, record the review, then commit and push intended changes.

Run focused and package-level verification.

Run focused and workspace verification, task-doc audit, diff hygiene checks, and post-commit dry-run release verification.

Run focused app validation, Playwright proof when available, the active alignment-page audit, and `git diff --check`.

Run focused handoff audits and diff checks.

Run focused lifecycle tests, the full package Node suite, canary `build:check`, package verification, task-doc audit, and `git diff --check`.

Run focused lifecycle verification, available build checks, task-doc audit, and diff hygiene checks.

Run focused Node test files.

Run focused package tests and package manifest/build checks.

Run focused package tests and whitespace checks.

Run focused package tests plus diff hygiene, then document the review and ship.

Run focused publish recovery tests, workspace node tests, task-doc audit, and diff hygiene checks.

Run focused publish-script tests, task-doc audit, and diff hygiene checks.

Run focused regression and archive/diff checks.

Run focused route text checks and `./scripts/skill-install-routing-audit.sh`.

Run focused search commands and `git diff --check`.

Run focused skill metadata and route checks.

Run focused static audits, perform manual mirror/version behavior checks, document results, commit, and push on the primary branch.

Run focused tests and archive/diff verification.

Run focused tests and hygiene checks.

Run focused tests, archive/generator audits, task-doc audit, and diff hygiene checks.

Run focused tests, workspace node tests, task-doc audit, and diff hygiene checks.

Run focused tests.

Run focused verification and record results.

Run focused verification and record the result in `tasks/todo.md`.

Run focused verification, generator check, archive audit, and relevant layer1 tests.

Run focused verification, update task review, then ship the intended changes.

Run full verification and fix any failures.

Run generated alignment bundle, layer1, archive/version/dependency, showcase, and whitespace checks.

Run generated bundle drift checks, skill hygiene checks, focused tests, showcase refresh if needed, and whitespace checks.

Run generated bundle drift, active `type: research` staged-workflow audit, targeted marker scans, archive/changelog checks, and whitespace checks.

Run generated-bundle drift checks.

Run generated-data parsing, alignment-page audit, targeted count scans, and whitespace checks.

Run generator and focused test verification.

Run generator checks, archive audits, focused tests, whitespace checks, and status checks.

Run generator drift checks and focused Vitest.

Run generator, archive, focused test, diff, and status verification.

Run package docs tests, package staging checks, dry-run release checks, and whitespace checks.

Run package dry-run and temp consumer install checks.

Run package staging and tarball dry-run inspection.

Run package syntax/package tests, alignment audit, whitespace checks, and ship-boundary review.

Run package tests and manifest/build checks.

Run package tests and package staging checks.

Run package-level verification and task-doc checks.

Run package, task-doc, diff, and deck inspections, including desktop and mobile viewport-fit checks when feasible.

Run pre-commit build and diff checks.

Run publish-prep verification:

Run publish-script syntax checks, focused tests, package build/check commands as needed, task-doc audit, and diff hygiene checks.

Run published canary smoke checks from disposable temp directories.

Run relevant documentation and alignment validators.

Run requested verification commands.

Run required validation, inspect generated text, document results, commit, and push on the primary branch.

Run required validation, inspect warnings, document results, create a ship manifest, commit, and push on the primary branch.

Run required verification commands and fix failures.

Run required verification, document results, commit, and push on the primary branch.

Run required verification, update `tasks/lessons.md`, document results, commit, and push on the primary branch.

Run routing, archive, generated bundle, package, and pair-app sanity checks.

Run skill integrity checks, generated drift checks, benchmark coverage, tests, targeted `rg` route checks, and `git diff --check`.

Run skill validation, mirror/version/archive parity, dependency, direct-primary policy, task-doc, catalog, package-manifest, and targeted shipping-route tests.

Run skill/version, archive, mirror, task-doc, and diff hygiene verification.

Run stale deck wording scans and `git diff --check`.

Run standard skill validation and benchmark coverage.

Run structural skill audits covering version metadata, broken dependencies, next-step routing, archive history, base parity, and mirror parity.

Run syntax, layer1, active-page audit, package/build boundary, and diff verification.

Run targeted active-skill route scans.

Run targeted convention, manifest, export, package, task-doc, and diff checks.

Run targeted drift scans for stale install wording, missing helper path, retired executable routes, and historical npm page labeling.

Run targeted duplicate audit, archive/version audit if skills change, diff checks, and status checks.

Run targeted existing repository checks appropriate for package metadata.

Run targeted package-boundary and publish-recovery tests.

Run targeted static checks and `git diff --check`.

Run targeted string/structure checks and `git diff --check`.

Run targeted tests and validation gates.

Run targeted tests, generator checks, grep audits, diff checks, and status checks.

Run targeted validation, archive/version audits, manifest/catalog checks, and diff hygiene.

Run targeted validation, record review/history, commit, and push.

Run targeted verification and `git diff --check`.

Run targeted verification and diff checks.

Run targeted verification and document results.

Run targeted verification and record the review.

Run targeted verification and update the task review.

Run targeted verification commands and `git diff --check`.

Run task-doc and generator verification, then commit and push the doc-only change.

Run task-doc audit and diff checks.

Run temp-repo parity tests against Node and bash paths before removing any fallback dependency.

Run text, interaction-harness, package, task-doc, diff, and opener verification.

Run textual, behavioral, package, task-doc, diff, and opener verification.

Run the canary publish dry run and publish both package names as `0.1.22-experimental.6` under `experimental` only.

Run the existing cross-pack routing audit.

Run the new script against the published npm package.

Run the package test script if available.

Run the package/build/test/readiness checks required by `package.json`, `packages/skillpacks/package.json`, `publish.sh`, and current touched surfaces.

Run the requested audit, generator, and focused layer1 verification; document results, commit, and push on the primary branch.

Run the requested dry-run/fixture verification, package tests, package build check, task-doc audit, and diff hygiene checks.

Run the requested generator, test, version, archive, dependency, showcase, and diff checks.

Run the requested helper, text, shell syntax, task-doc, and diff hygiene checks.

Run the requested verification commands.

Run the required verification for both repos.

Run the updated audit against Alignmeant and amend the active BIP page if it only contains the stale selected-channel drafting option.

Run verification and record results.

Run verification and record review results.

Run whitespace checks and review the diff.

Run whitespace checks.

Run whitespace/diff checks and focused convention verification.

Run workflow/routing quality audits covering install, pack, ship-end, research-loop, alignment, and researchish lifecycle contracts.

Scan changed pages for preserved titles/context, decisions, gates, assumptions, proposed file changes, and TTS script.

Search active skills for `$pack`, `pack install`, `scripts/pack.sh install`, `scripts/pack.sh refresh`, and missing-skill fallback references.

Search docs/package sources for user-facing pack-install guidance.

Seed the new repo with `README.md`, `tasks/next-step.yaml`, `research/youtube-owner-analytics-platform.md`, and `alignment/investigate-youtube-owner-analytics-platform.html`.

Separate active contracts from archives, generated data, and the `pack` skill implementation itself.

Separate active operating docs from generated artifacts, prompt history, archives, and historical reports.

Shared design-tree-loop convention matches the skill contract.

Smoke-check the published canary package names.

Snapshot and repair `apps/next-level-startup` as the pilot, preserving unrelated dirty files byte-for-byte.

Snapshot the remaining affected worktrees, approve the root dry-run only if it contains no unexpected/unmanaged deletion, run fleet refresh, and verify the formerly failing projects.

Snapshot, dry-run, repair, and verify `apps/next-level-startup` with published `.6`.

Specify stage-click navigation must not hijack slide content, links, buttons, form controls, filmstrip, topbar, footer, or review inputs.

Spot-check command examples and package-version wording.

Spot-check one research framework subskill and one product-design skill.

Stable package staging emits a stable manifest even after a canary manifest was generated in `dist`.

Stage extraction from `b7c0775bc` for showcase + benchmarks. **Reconciliation:** completed out-of-band on GitHub; verified read-only.

Stage only intended files.

State that `claude` and `codex` mirrors must be upgraded together where both exist.

Static audits enforce canary briefing-first wording in the relevant shared docs/assets.

Step 3.1: Node Project Config Parity.

Step 3.2: Pack Normalization And Alias Parity.

Step 3.3: Install/Remove/Refresh Parity.

Step 3.4: Locking And Drift Parity.

Step 3.5: Compatibility Closure.

Step 4.1: Documentation readiness.

Step 4.2: Documentation contracts.

Step 4.3: Dry-run release checks.

Step 4.4: Ship.

Step 5.1: Release preflight.

Step 5.1b: MIT metadata prepublish fix.

Step 5.2: External publish gate after metadata validation.

Step 5.3: Public npm publish.

Step 5.4: Published-package verification.

Step 5.5: Commit and push post-publish evidence.

Support `--json` machine-readable output for tests.

Support default human summary mode without file writes.

Target list is derived from active, non-archived `SKILL.md` files.

The `0.1.19` section summarizes YouTube meta research, rapid deck graduation reconciliation, Platform Fit Workshop, clean-context design-tree handoff wording, and refreshed package/catalog metadata.

The benchmark repo contains the benchmark harness/results at its repo root and points to `GeorgeQLe/agentic-skills-benchmarks`.

The briefing convention explicitly defines the dense-page-first, deck-open-first workflow.

The continuation command is still available in `## Invoke With YAML` / `agent_routing.command`.

The convention explicitly requires red borders while required gate questions are unanswered and green borders once answered.

The experimental publish dry run reaches its normal dry-run completion path.

The fleet recovery reports `60 ok, 0 flagged, 0 failed`, or any discovered count discrepancy is investigated and documented before proceeding.

The flow-tree schema and skill language use matching branch names for UI experiment/review nodes.

The HTML page preserves the research brief's executive recommendation, repo/skill needs, API capability matrix, OAuth/quota/delay findings, architecture, CLI/file contracts, normalized evidence contract, scheduling, security, failure handling, validation, open questions, and official source list as structured HTML.

The local new repo exists at `/Users/georgele/projects/tools/youtube-owner-analytics`, uses `master`, and has the correct `origin`.

The migration itself is delivered for human review through the new branch/PR workflow.

The new repo includes the copied research brief and alignment page at the same relative paths.

The npm publish target includes both templates and excludes denied repo paths.

The page declares `data-alignment-category="utility"`, `data-visual-tier="document"`, and `data-alignment-status="review"` and includes the required TTS and question-nav script tags.

The pilot and full fleet contain no package-managed dangling links after recovery, with application repositories left uncommitted and existing dirty content preserved.

The published `.5` canary no longer reports convention-doc drift after installing and removing a deck-backed pack.

The second fleet dry-run proposed zero installs, updates, or removals, and the ten formerly failing projects contained zero dangling links.

The Showcase repo contains the app at its repo root and points to `GeorgeQLe/agentic-skills-showcase`.

The skills repo preserves the current `agentic-skills` history and remote.

The static briefing-slide audit fails valid-looking decks that include required gate questions without the required border/status affordance.

The three safety subskills are base skills available to both Claude and Codex and refuse force-push, silent merge, duplicate issue/PR creation, destructive unmerged-branch deletion, and ambiguous dirty-tree adoption.

Three independent repos exist for skills, showcase, and benchmarks — GitHub authoritative and populated; skills worked from this single local checkout, showcase/benchmarks remote-only by choice (no accidental nested local repos).

Tighten `docs/alignment-page-convention.md` so the initial BIP channel-selection page requires only target-channel selection until approved channel-selection YAML is consumed.

Tighten `logic-wiring` routing so prototype approval routes only to UAT; consolidation remains a later evidence/scope decision.

Tighten UAT variant-evaluation readiness language to distinguish built/evaluated, built/not-run, approved-unbuilt/deferred, and explicitly MVP-excluded variants.

Trace `uninstall-global` ownership detection in the CLI source.

Trace CLI ownership/removal logic for `uninstall-global`, including current-package source checks.

Track base installation in `.agents/project.json` so `skillpacks refresh` rebuilds base skills from the current package snapshot.

Treat `--all` as the current default recursive cleanup from the command cwd.

Treat `--global` as user-home cleanup scope for project config cleanup, deprecated alias cleanup, and optional base-skill reinstall migration.

Unmanaged dangling links and valid pinned links remain untouched.

Update `alignment/index.html` metadata if confirmation state/date changes require it.

Update `alignment/index.html` title/search metadata for the amended pages.

Update `alignment/index.html` to mark the entry confirmed.

Update `alignment/index.html` with the new dated page entry in the Utility & Maintenance category.

Update `alignment/index.html` with the new review page entry.

Update `CHANGELOG.md` `0.1.20` to cover explicit alignment gate metadata, session-triage evidence-path fixes, UX variations assemble-stop handling, repo-local skillpacks install prohibition, and regenerated catalog/package metadata.

Update `CHANGELOG.md` for `v0.16`.

Update `consolidate-prototypes` mirrors so missing UAT evidence routes with plain install-then-run guidance.

Update `create-alignment-page` and convention guidance to use the scaffold as infrastructure without routing normal skills through the skill.

Update `design/flow-tree-sample.yaml` with a realistic platform-fit matrix and one thin platform probe.

Update `design/flow-tree.schema.json` to `v0.5` with additive `platform_fit` and `prototype_build_item.platform_probe` support.

Update `docs/alignment-page-convention.md` so BIP pages rank target channels, assign `recommended` / `optional` / `not-now`, preselect recommended channels for user confirmation, and treat the submitted gate as channel-selection approval only.

Update `docs/alignment-page-convention.md` so post-confirmation BIP pages live at `alignment/bip/{skill-name}.html`.

Update `docs/alignment-page-convention.md` so Stage 1 is minimal scope discovery, not synthesized research.

Update `docs/alignment-page-convention.md` with explicit `data-approval-effect` requirements and compile-path validation fixtures.

Update `docs/briefing-slides-convention.md` so required keyboard navigation includes `A` and `D`.

Update `docs/decks.md`, `packs/vard/PACK.md`, `packs/ord/PACK.md`, `packs/devtool/PACK.md`, `docs/operating-modes.md`, `docs/skills-reference.md`, and stale pack workflow references.

Update `docs/interrogation-page-convention.md` inside the canonical `interrogation-convention` block.

Update `docs/release-runbook.md`, `docs/skillpacks-npm-distribution.md`, and `CHANGELOG.md` with canary release commands, validation, and publishing rules.

Update `docs/skillpacks-npm-distribution.md` so deck installation is modeled as manifest metadata, not opaque monolith presets.

Update `logic-wiring` mirrors with explicit Pack Availability Guard wording and non-bare UAT handoffs.

Update `PACK.md`, the YouTube router if needed, and benchmark coverage fixtures.

Update `packs/base/PACK.md`, skill metadata, UI metadata, public references, and the packaged manifest so the new subskills are universally available.

Update `publish.sh --current` so both-package-already-published state skips publishing and reruns final published-package verification plus post-publish source-state instructions.

Update `scripts/audit-alignment-pages.mjs` and focused tests to enforce the new BIP path shape and clarity fields.

Update `ship` and `ship-end` contracts for terminal-only absent-setting prompting, explicit-false suppression, and no BIP blockers/review gates.

Update `sync.md` so `/sync` installs `publish-canary` onto `~/.local/bin`.

Update active `SKILL.md` install guidance; keep `/pack`, `$pack`, and `scripts/pack.sh` only as explicit legacy/source-checkout behavior.

Update active Pattern A docs, parent orchestrators, framework subskills, and audits to use YAML-only review-gate continuation.

Update active task docs with this implementation contract.

Update base pack metadata so the new skill is discoverable in the packaged canary.

Update both mirrors so high-confidence deck handoffs require copy-pasteable post-install command lines as secondary context.

Update Business AFPS deck docs to name customer discovery as the default first workflow after install.

Update canonical `docs/design-tree-loop-convention.md` handoff wording.

Update canonical design-tree docs and regenerate bundled `DESIGN-TREE-LOOP.md` files.

Update canonical design-tree loop convention and regenerate generated bundles.

Update canonical JavaScript/YAML guidance to apply and compile the hidden agent payload while retaining backward-compatible fallback behavior.

Update catalog, package, lifecycle, ownership, audit, generator, and docs code so `packs/base` remains `scope: "base"` with `pack: null`.

Update CLI help and parser for `cleanup [--all|--global] [--reinstall-base] [--dry-run]`.

Update Codex and Claude `create-briefing-slides` mirrors to declare and reference the shared convention.

Update Codex and Claude `idea-scope-brief` instructions so the one-time BIP prompt enables `set-bip on`, dismisses the prompt, and generates/opens `alignment/bip-idea-scope-brief.html` after `idea-brief.md` is written.

Update Codex and Claude `uat`, `logic-wiring`, and `consolidate-prototypes` contracts to run handoff verification immediately before terminal handoff text or `agent_routing`.

Update Codex and Claude `uat`, `logic-wiring`, and `consolidate-prototypes` contracts.

Update commands, comments, and documentation to identify these files as website-owned.

Update create-alignment-page and convention text for scaffold usage.

Update design-tree convention, alignment-page routing, consumers, metadata, catalog/manifest outputs, and generated bundles.

Update docs/help to describe cleanup and the experimental release policy for future canary-only features.

Update focused layer1 assertions for the new contract.

Update focused layer1 assertions for unified response compilation, local feedback visibility, generated bundle wording, and upgrade-page feature wording.

Update focused layer1 coverage for canonical and generated bundle text.

Update focused layer1 coverage for the new agent-owned-depth and unknown-routing contract.

Update focused layer1 tests for scope-approval-before-research behavior.

Update focused manifest/lifecycle/package-boundary/layer1 tests for the new canonical paths and legacy marker compatibility.

Update focused tests for source resolution, packaged asset resolution, stub validation, legacy fallback, and metadata interpretation.

Update focused tests or audits that cover research-roadmap/reconcile-research/Pattern A next-step routing language.

Update help/docs without changing global install semantics.

Update layer1 fixtures for linked-BIP failure, linked-BIP read-only preview success, and approved-BIP active final approval success.

Update lessons and task review notes with the correction pattern and verification results.

Update live Skills Showcase research references from four decks to five decks.

Update mirrored state-model `SKILL.md` handoff wording if regeneration does not cover it.

Update npm distribution docs and quickstart wording to explain base versus global install paths.

Update only the alignment page styles/markup needed to keep the chart within the viewport.

Update ownership checks, install markers, manifest generation, package build staging, docs, and tests to recognize `packs/base`.

Update pack workflow docs, skill next-step contracts, skills reference, product-design `PACK.md`, and global skill browser mappings.

Update package lifecycle tests, product-design route tests, and manifest/catalog expectations for replacement skills only.

Update package staging and npm boundary assertions.

Update package-included docs with npm usage and git-checkout usage.

Update positioning, UI interview, UX variations, prototype, spec interview, and research-roadmap routing.

Update quality-gate, operating-mode, invocation-type, skill-reference, pack, and next-step documentation for the new lifecycle.

Update README/help/changelog with cleanup wording and canary-only experimental-feature guidance.

Update refresh planning/dry-run behavior so source-path-only drift is not counted as a proposed update.

Update refresh/install lifecycle outcomes so content-visible changes are separated from marker-only metadata maintenance.

Update report template and final response requirements to include the URL record path and the three Test and Compare pairs.

Update review notes and history.

Update review notes, commit, and push intended changes only.

Update review notes, history, commit, and push intended changes.

Update root generated blocks, provisioner templates, package docs, and routing contract to standardize on `npx skillpacks install <pack-or-skill>`.

Update routing audit so active `/pack install` and `$pack install` guidance fails unless allowlisted.

Update setup/troubleshooting/script docs so track-latest installs are described as managed copies/directories and pinned archives are the symlink case.

Update shared BIP policy so `idea-scope-brief` is the only BIP gate-question owner and all other BIP output is read-only/non-blocking.

Update shared reusable routing docs for UAT-evidence plus explicit unbuilt/deferred-branch handling.

Update shared routing docs where they define reusable readiness rules.

Update social post/video output-shape guidance where needed so generated candidates carry audience/context fields.

Update social routing conventions so enabled BIP mode loads every bundled channel convention, with `alignment.bip_platforms` used only as prioritization metadata.

Update stale count references in current Skills Showcase docs and indexed alignment pages, including `tasks/pack-card-hierarchy.md`, `alignment/skillmap.html`, `apps/skills-showcase/docs/deck-builder-ux.md`, `research/skills-showcase/idea-brief.md`, `research/skills-showcase/idea-brief-interview.md`, and `alignment/idea-scope-brief-skills-showcase.html` as confirmed by fresh scans.

Update task/history notes and manifest.

Update the 14 global files listed in the audit: `pack`, `skills`, `init-agentic-skills`, `provision-agentic-config`, `afps-status`, `codebase-status`, and `idea-scope-brief` for both Claude and Codex where present.

Update the alignment index if needed, verify the page, update `.session-insights/`, and record the review in `tasks/todo.md`.

Update the canonical alignment-page convention so BIP is a post-confirmation artifact at `alignment/bip-{skill-name}.html`, archived before replacement, indexed, and opened after the skill concludes.

Update the canonical alignment-page convention to allow linked-BIP handoff text and read-only final-approval preview content, while blocking active final artifact approval controls until BIP is approved or narrowly not-applicable.

Update the canonical briefing-slides convention with fit-to-slide authoring, linked-reference, and viewport overflow verification rules.

Update the canonical Browser open paragraph in `docs/alignment-page-convention.md`.

Update the canonical convention so visible `data-recommended-answer` text is separate from hidden `data-agent-recommended-answer` payload text.

Update the canonical interrogation convention.

Update the central alignment index and task review notes.

Update the design-tree convention and mirrored `ux-variations` handoff wording.

Update the live deck documentation sentence that referenced npm install presets.

Update the npm distribution page's deck-installation section and gate to include Game AFPS.

Update the provisioned `AGENTS.md`/`CLAUDE.md` blocks and this repository's active copies to make issue-backed branch/PR delivery the default.

Update the shared briefing-slides convention to allow slide-scoped bottom-bar feedback sidebars.

Update the Skills Showcase deck-builder page to use five canonical decks.

Update the task docs with the active implementation contract and review section.

Update the website generators and package staging script to read through this internal catalog layer.

Update the workflow-design page to the five-deck model: VARD, ORD, Business AFPS, Devtool AFPS, and Game AFPS.

Update VARD/ORD ship and traction skill handoffs in Claude and Codex mirrors, bump versions, and update changelogs.

Update visible dogfood deck text to demonstrate click-stage and `A`/`D` navigation.

Updated and versioned Codex `plan-phase`, `exec`, `expert-review`, `ship`, and `ship-end`; Claude sources remain unchanged.

Use `/tmp/skillpacks-npm-cache` by default and isolated `mktemp` project directories.

Use that registry in Node lifecycle install/init/refresh flows to overwrite only `.agents/skillpacks/docs/**`.

Use that registry in package staging so npm installs contain the same canonical docs.

Validate source package/manifest metadata with canary package staging.

Validate the `exec` alias/skill precedence claim against the current manifest.

Validate the current BIP target-channel language in the canonical convention, generated bundles, tests, and recent history.

Validate the current convention, audit, fixture coverage, and git history for required-gate slide status handling.

Validate the current dirty paths, exact release metadata diffs, and whether they are pre-existing `0.1.16` work.

Validate the proposed outputs against current `youtube-video-audit` owner-analytics inputs and requested schedule/failure scenarios.

Validate the source package/manifest metadata and canary package staging boundary.

Validate the user claim against the alignment page and recent deck documentation.

Validate the user's claim against the current `ship-end` skill contract, source pack mirrors, and recent history.

Validate whether `refresh` touches global/user-home skills.

VARD graduation routes to `npx skillpacks install business-research`, then `$idea-scope-brief` for raw/new framing or `$customer-discovery` for clear shipped concepts with traction evidence.

Verification commands pass, or any residual failures are documented as unrelated pre-existing issues.

Verification commands show the new wording and no generator drift.

Verification demonstrates the focused regression passes.

Verification output is recorded in the Review section.

Verification results are captured in `CHANGELOG.md` and `tasks/todo.md`.

Verify (acceptance): confirmed on the refreshed runtime `.codex/skills/ux-variations/{SKILL,DESIGN-TREE-LOOP}.md` — SKILL.md is v0.33 with the §0c exception + step-8 carve-out (lead with HTML review path, only the compiled response YAML, no repeat-`$ux-variations`/second-YAML/`Exact next command`, retains completed-count + durable-cursor); DESIGN-TREE-LOOP.md carries the Assemble-ready review-gate exception.

Verify `list`, pack install, individual skill install, deck install, and `doctor`.

Verify `node bin/skillpacks.mjs list`.

Verify `npx skillpacks@latest list` against npm.

Verify archive snapshots exist for all three edited pages.

Verify deck text, repo-local reference links, diff hygiene, and opener behavior.

Verify each edited page contains its dated addendum and required key findings.

Verify each repo. **Reconciliation:** verified read-only via `gh repo view` and `gh api .../contents`.

Verify every active page has a matching archive copy.

Verify final compiled YAML is complete and has no negative section feedback.

Verify no `specs/skills-showcase/user-flow-deck-creation*` files are modified.

Verify pack, individual skill, and deck-backed pack removal.

Verify package metadata and manifest metadata both report `0.1.22-experimental.1` / `canary`.

Verify skill-level pin to `v0.0`, unpin back to latest, and unsupported direct `install name@version` syntax.

Verify source audit, packaged audit, focused package tests, dry-run package contents, task docs, and diff hygiene.

Verify temp consumer repo install/status/doctor behavior.

Verify the alignment page visually in the browser.

Verify the alignment-page audit passes for the active page set.

Verify the design doc contains the revised deck approval, manifest resolver, package-list fields, and registry-tag fields.

Verify the doc contains the approved decisions, phase roadmap, and no implementation drift.

Verify the git-checkout install path remains functional.

Verify the global and project-local `codebase-status` skill markers and versions.

Verify the proposed canonical markdown files remain unwritten.

Verify UI canonical files remain unwritten before final UI approval.

Verify: `node scripts/upgrade-design-tree-loop.mjs --check` and `node scripts/audit-alignment-pages.mjs` both exit 0. (Also ran `node scripts/upgrade-alignment-page.mjs --check` → exit 0, and `git diff --check` → clean.)

Verify: contract grep — `rg -n "assemble-ready|Progress Handoff Block|whole-set alignment review|Invoke With YAML" packs/product-design/codex/ux-variations/SKILL.md docs/design-tree-loop-convention.md` confirms assemble-ready (SKILL.md:285, convention §5:476–481) decoupled from repeat-command/second-YAML framing; setup (234–235) + per-variation (242) stops retain it. Canonical-source acceptance replay confirmed (a) leads with HTML review path, (b) only compiled response YAML, (c) no repeat-`$ux-variations`/second-YAML/`Exact next command`, (d) retains completed-count + durable-cursor.

Version bump (codex v0.32→v0.33, claude v0.31→v0.32; variants had diverged so each bumped one decimal step) — ran `scripts/skill-archive.sh` for both (archived codex v0.32, claude v0.31), bumped `version:` frontmatter, added `CHANGELOG.md` entries. Original path was `scripts/skill-archive.sh packs/product-design/codex/ux-variations` and `scripts/skill-archive.sh packs/product-design/claude/ux-variations`, bump `version:` frontmatter in both, add `CHANGELOG.md` entries describing the assemble-stop review-gate carve-out — behavioral change requires archive + changelog.

Write `apps/skills-showcase/docs/animation-plan-deck-builder.md` from the approved page content and gate answers.

Write `docs/skillpacks-npm-distribution.md`.

Write `research/devtool-docs-audit.md` findings-first.

Write and verify `research/youtube-owner-analytics-platform.md`.

Write canonical user-flow spec and interview log under `specs/skills-showcase/`.

Write findings-first documentation audit output.

Write prompt-history records for the `skill-creator` and `ui-interview` invocations.

Write the draft only to `research/skills-showcase/_working/preliminary-ui-interview-research.md`.

Write the findings and remediation order to `research/skillpack-cli-routing-audit.md`.

Competitive Intel

Research Hub

Roadmap

Timeline