feat: v0.8.0 — classifier QA fix, routing trace, cold-start UX (#39-#46)#49
Merged
SmartBrandStrategies merged 5 commits intomainfrom Mar 6, 2026
Merged
Conversation
…versal Closes #43: Remove shell: true from runGit() in git-helpers.ts. Node.js resolves the git binary via PATH directly without a shell on WSL, Linux, macOS, and Windows. shell: true is unnecessary and allows shell metacharacters in args to be interpreted as shell syntax. Closes #42: Validate module paths in adf create before path.join. Paths containing ".." or absolute paths are rejected with a clear error. A secondary resolved-path check confirms the final path stays within the .ai/ directory, guarding against platform-specific bypass patterns. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…ld section charter adf populate [--dry-run] [--force] [--ai-dir <dir>] Reads package.json, README.md, and stack detection signals to auto-fill ADF files with project-specific content after charter adf init. Populates CONTEXT in core/backend/frontend.adf and STATE in state.adf. Idempotent: skips files with non-scaffold content unless --force. patcher: unknown ops now produce a clear error listing valid op names instead of the cryptic "handlers[op.op] is not a function" TypeError. CORE_SCAFFOLD: add a CONTEXT section placeholder so ADD_BULLET section:CONTEXT works immediately after adf init without requiring ADD_SECTION first. adf patch help: list all valid ops with concrete usage examples. bootstrap/init next steps: point to charter adf populate as step 1 instead of generic "edit core.adf manually" guidance. harness: extend SDLC corpus and runner with mixed QA/backend signal scenarios that exposed the classifier routing issues filed in #44/#45. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
#45 #46) - QA compound phrases (smoke test, contract test, schema compat, approval gate, verified against, test fixtures) now fire a phrase-level override BEFORE keyword scoring so they cannot be outvoted by raw infra/backend keyword count (#44, #45) - contentToModule() returns { module, phraseOverride?, scores } so all per-module candidate scores are visible to callers (#46) - ClassificationResult gains optional routingTrace?: RoutingTrace with headingModule, phraseOverride, and candidateScores for debugging - adf tidy --verbose prints per-item routing rationale (module, section, trigger scores or phrase override) (#46) - 9 new tests covering phrase override routing, guard against absent qa.adf in triggerMap, and routingTrace shape (#44 #45 #46) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…rse-pointer warn (#39 #40 #41) - adf migrate --keep-summary: injects auto-generated "## Architecture Summary" block listing migrated module names and section counts so CLAUDE.md thin pointer gives agents architectural orientation without duplicating rules (#39) - adf migrate --audit: prints per-module breakdown (constraints/context/ advisory counts) and flags potential misroutes via routingTrace — items routed to core.adf that scored non-zero for a specialized module (#40) - doctor: adds 'INFO' status tier (soft, does not fail overall check); warns [info] when thin-pointer CLAUDE.md has <15 lines and no stack/framework keywords — agents have zero orientation, suggests charter adf populate (#41) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <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
2 participants
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
This PR lands all remaining issues for v0.8.0 across three clusters:
Classifier cluster (#44 #45 #46)
smoke test,contract test,schema compat,approval gate,verified against,test fixturesnow fire a phrase-level pre-routing override before keyword scoring — prevents infra/backend keyword count from outvoting compound QA signalsClassificationResultgains optionalroutingTrace?: RoutingTracewithheadingModule,phraseOverride, and per-modulecandidateScoresfor debuggingadf tidy --verbose: prints per-item routing rationale (module, section, trigger scores or phrase override phrase)Cold-start cluster (#39 #40 #41)
adf migrate --keep-summary: injects auto-generated## Architecture Summaryinto the thin-pointer CLAUDE.md listing which modules content went to (constraints/context/advisory counts) — agents get architectural orientation without rule duplicationadf migrate --audit: prints per-module breakdown post-migration and flags potential misroutes (items routed tocore.adfthat scored non-zero for a specialized module viaroutingTrace)doctorsparse-pointer[info]warn: newINFOstatus tier (soft, does not fail--ci); fires when a thin-pointer CLAUDE.md has <15 lines and no stack/framework keywords, suggestscharter adf populateAlso includes (from feat/adf-populate)
charter adf populateauto-fills ADF files from codebase signals (package.json, README, stack detection)CORE_SCAFFOLDincludes CONTEXT placeholderadf create,shell: trueremoved fromrunGit()Test plan
pnpm test— 273+ passing (1 pre-commit hook test is a WSL timing flake when run in full suite, passes in isolation)charter adf migrate --dry-run --keep-summaryon a real CLAUDE.mdcharter adf migrate --auditshows per-module breakdowncharter doctorshows[info]on sparse thin pointerscharter adf tidy --verboseshows routing trace per itemCloses #39, #40, #41, #44, #45, #46
🤖 Generated with Claude Code