Repository navigation
P0 learning quality: learn lint + ADR 0001 validation milestone (T11, T12) - #45
Merged
Merged
Conversation
- lintLearnings scans a store for duplicate LEARN IDs, near-duplicate rules (token Jaccard >= 0.4), and learnings with no checkable trigger - learn lint CLI: pretty/json output, exit 1 on findings (CI-usable) - On this repo's real store it surfaces 12 duplicate-ID collisions, 4 near-duplicates, and 2 no-trigger learnings Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…pus) One-shot consolidation pass over the real store (55 files, 53 distinct learnings): 6 candidate topic pages with provenance and trigger tables, contradictions + disposition table, 12 mechanical compilation targets. Verdict: acceptance criteria met (92% grouped, contradictions non-empty); ADR 0001 proceeds toward Accepted with one revision — topic pages own a trigger table, skills are the delivery vehicle. Key evidence: the store already contained a failed ad-hoc consolidation (short-slug generation) that collided IDs 001-012 because merge/retire don't exist as verbs. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…ation - Move the dispositions doc out of wiki-candidates: it is ADR 0001's change set + review queue, not wiki content — docs/consolidation/ now holds dated change sets - T13 (sheal consolidate: emit a reviewable change set): productize the pass T11 ran manually - T14 (compile mechanical learnings into hooks and checkers): emitter v0 for the change set's 12 mechanical targets Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
… in .sheal, snapshots publish on promotion) - docs/wiki-candidates/ and docs/consolidation/ collapse into docs/adr/0001-validation/ — they are evidence for the ADR's acceptance, not living data; recurring output belongs to sheal under .sheal/ (invariant: nothing in docs/ is something sheal regenerates) - Q6 answered (hybrid): working wiki pages live in .sheal/wiki/, reviewed snapshots are published into git on promotion — the git snapshot is an emitter target, like compiled skills Co-Authored-By: Claude Fable 5 <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.
Two commits advancing the learning-quality track:
T12 (Support
sheal learn lint: corpus hygiene checks)lintLearningsscans a learnings store for duplicate LEARN IDs, near-duplicate rules (token Jaccard ≥ 0.4), and learnings with no checkable trigger.sheal learn lintCLI: pretty/JSON output, exit 1 on findings (CI-usable),--globalfor the user store.test/learn-lint.test.ts(5 tests) written first and red, then the implementation; includes an end-to-end CLI test.T11 (Run the ADR 0001 validation milestone: consolidate the LEARN corpus)
One-shot consolidation pass over the real store (55 files, 53 distinct learnings — the store had silently diverged from AGENTS.md's 30). Output in
docs/wiki-candidates/:Nothing in
.sheal/learnings/was modified — the wiki-candidates are a reviewable proposal; executing the disposition table is a follow-up task.Testing
Full suite 319/319 green, lint 0 errors,
sheal learn lintdemonstrated on the real corpus.Task trail: docs/tasks/t11-learning-corpus-consolidation-pass.md, docs/tasks/t12-learn-lint-corpus-hygiene.md
🤖 Generated with Claude Code