Phase 15 — KB index analysis caches#280
Open
leeovery wants to merge 11 commits into
Open
Conversation
Add 'analysis' to INDEXED_PHASES, extend deriveIdentity to parse
.workflows/{wu}/.state/{research-analysis,gap-analysis}.md paths with
filename whitelist enforcement, extend discoverArtifacts to scan caches
by file existence. Mirror research.json as analysis.json chunking config
(confidence: low). Rebuild bundle and cover with CLI + integration tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Append a `knowledge index` call to section E after the manifest writes so the analysis content surfaces in future contextual queries. Mirrors the surface-and-continue error pattern from import-files.md. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Append a `knowledge index` call to section E after the manifest writes so the gap-analysis content surfaces in future contextual queries. Mirrors the surface-and-continue error pattern from import-files.md. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…ia self-healing Both skills load self-healing references that now invoke `knowledge index` on analysis cache writes. Declare the tool preemptively so the bridge and refinement-session flows don't hit a permission prompt mid-stream. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The for-INDEXED_PHASES loop in discoverArtifacts would no-op on the new 'analysis' value (no manifest.phases.analysis exists), but the silent fall-through is unclear. Make the skip explicit, matching the imports special-case pattern. Behaviour is identical. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three review-driven fixes: - SKILL.md "What is indexed" / "rebuild" prose was a hardcoded enumeration, not auto-derived — add the analysis row. - reindex-work-unit.md only walked manifest-backed phases plus imports; reactivating a cancelled work unit left its analysis caches indexable on disk but absent from the KB. New section C probes the two cache paths and re-indexes any present. - Surface the .state operational-file exclusion in "What is NOT indexed". Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The analysis writes its cache to .state/discussion-gap-analysis.md (prefixed), but the indexing whitelist accepted only gap-analysis.md (no prefix). The reference's `knowledge index ... discussion-gap-analysis.md` call would have thrown UserError at runtime, and bulk discovery would have looked at the wrong path and never found the file. Tests passed pre-fix because they exercised the whitelist with the wrong filename — a self-confirming gap. Updated tests to use the canonical filename so the bug can't reintroduce. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two further doc-accuracy gaps surfaced during review: - Confidence-tier table listed only research + imports under 'low'; analysis sits in the same tier and consumers need to distinguish via the provenance line's phase field. - 'index' command's path-derivation note covered the standard and imports shapes but not '.state/' analysis paths. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…sertions - reindex-work-unit.md section B's 'no imports' branch returned to caller, bypassing my new section C. An epic with no imports but with on-disk analysis caches would not have them re-indexed on reactivation. Route both B branches to C instead. - CLI tests 90 and 91 grep for 'analysis' in query output — but the Source: line contains the .state filename, which also matches. Switch to anchored '^[analysis |' assertions so the test actually exercises the provenance line. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The [ -f file ] && cmd shortcut exits non-zero when the file is absent (the test command's exit code), which Claude's Bash tool surfaces as an error and aborts the reference. A fresh epic with no analysis caches yet would always trip this. Switch to if/then/fi so the absent-file branch exits 0 cleanly. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
.workflows/{wu}/.state/{research-analysis,gap-analysis}.mdcache files into the knowledge base under a newanalysisphase value, so the full reasoning behind surfaced topics is searchable alongside research/discussion/spec content.knowledge remove --work-unitpath.research-analysis,gap-analysis) prevents operational.state/files (migrations, environment-setup) from ever entering the KB.Changes
src/knowledge/index.js: extendINDEXED_PHASES,deriveIdentity(separate.state/match with whitelist), anddiscoverArtifacts(file-existence scan inside the per-work-unit loop, after imports).skills/workflow-knowledge/scripts/knowledge.cjs: rebuilt bundle.skills/workflow-knowledge/chunking/analysis.json: mirrorsresearch.jsonwithconfidence: "low".skills/workflow-shared/references/{research-analysis,discussion-gap-analysis}.md: appendknowledge indexcall after manifest cache-metadata writes in section E (surface-and-continue error pattern).skills/workflow-{inception-process,bridge}/SKILL.md: declareBash(node .claude/skills/workflow-knowledge/scripts/knowledge.cjs)so self-healing-driven indexing doesn't prompt.tests/scripts/test-knowledge-cli.sh) and 2 new integration cases (tests/scripts/test-knowledge-integration.cjs) covering happy path, whitelist enforcement, idempotency, bulk discovery, cancelled-wu skip, and path-traversal guards.Catch-up for existing work units
Existing in-progress epics with stale caches will re-index naturally on the next analysis re-run (Phase 7's stale-cache trigger). Users wanting immediate searchability of an unchanged cache can run
knowledge index .workflows/{wu}/.state/{filename}.mdmanually.Test plan
node --test tests/scripts/test-*.cjs— 641 tests passbash tests/scripts/test-knowledge-cli.sh— 215 pass (12 new)bash tests/scripts/test-knowledge-build.sh— bundle freshness verifiednpm run build) produces a bundle identical to the committed one🤖 Generated with Claude Code