P5a shared kit extraction inventory

[andrei-hasna]

Summary

• Add docs/SHARED_KIT_EXTRACTION_INVENTORY.md with the P5 shared-kit extraction inventory and package-boundary proposal.
• Classify storage-kit, runtime/path, CLI, events, and logging/diagnostics candidates by automation readiness and approval needs.
• Record the narrow follow-up todo 2b166a36-1e7c-4882-be7f-610c4f478d39 for OpenLoops storage-kit normalization.

Validation

• git diff --check
• git diff --cached --check
• staged credential-pattern scan: no matches
• bun run typecheck
• bun run test:boundary

Notes

• Docs-only change; no package metadata, release-age exclusions, publishing, or consumer migrations.
• Commit/push hooks printed Can't find lefthook in PATH, but commit and push completed.

[andrei-hasna]

Worker review is blocking merge for PR #80 for now.

Blockers found:

• The follow-up section is stale: docs/SHARED_KIT_EXTRACTION_INVENTORY.md says task 2b166a36 is automation-ready/safe to auto-route, but live todos state shows that task is already completed and produced PR Normalize OpenLoops storage-kit snapshot #84. Update the section to say it was completed/superseded by PR Normalize OpenLoops storage-kit snapshot #84, or close P5a shared kit extraction inventory #80 in favor of Normalize OpenLoops storage-kit snapshot #84.
• The generator guidance is ambiguous: the inventory repeatedly says to use the current/latest @hasna/contracts vendor-kit and suggests bunx @hasna/contracts vendor-kit --check, but the PR head has @hasna/contracts ^0.4.1, hasna.contract.json kit 0.4.1, vendored storage-kit manifest 0.4.0, and newer machine-level contracts output exists. Pin the exact generator source/version/ref expected for follow-up workers.
• Validation on current main plus PR P5a shared kit extraction inventory #80: git diff --check passed, bun run typecheck passed, bun run test:boundary passed, but full bun test ended with 727 pass / 16 skip / 1 timeout failure in loops CLI > todos task provider rules fall back to fixed Codewith pools and reject invalid hints.
• GitHub review gate is still unmet: latestReviews is empty and reviewDecision is blank. The active gh account is the PR author, so a non-author review is still required.

I did not merge. Worktree was restored clean after validation.