Skip to content

docs: add dogfood report for v2.2.3-dev#83

Merged
carlos-alm merged 1 commit intomainfrom
docs/dogfood-report-v2.2.3-dev
Feb 24, 2026
Merged

docs: add dogfood report for v2.2.3-dev#83
carlos-alm merged 1 commit intomainfrom
docs/dogfood-report-v2.2.3-dev

Conversation

@carlos-alm
Copy link
Copy Markdown
Contributor

Summary

  • Adds the dogfood report (generated/DOGFOOD_REPORT_v2.2.3-dev.44e8146.md) that was left behind as an untracked file after the dogfooding session
  • Updates the dogfood skill (.claude/skills/dogfood/SKILL.md) with a new Phase 10 — Commit the Report so future reports are always committed — either with the first bug-fix PR or in a dedicated PR

Test plan

  • Report file is present in generated/
  • Skill update adds clear instructions for committing reports

…ports

- Add the dogfood report that was left as an untracked file
- Add Phase 10 to the dogfood skill ensuring future reports are
  committed with the first PR or in a dedicated PR
@claude
Copy link
Copy Markdown

claude bot commented Feb 24, 2026

Claude finished @carlos-alm's task —— View job


PR Review: docs: add dogfood report for v2.2.3-dev

Todo List:

  • Phase 1: Root Cause Analysis - understand what problem this PR solves
  • Phase 2: Configuration Change Detection - check for high-risk changes
  • Phase 3: Test Integrity - look for red flags in testing
  • Phase 4: Code Quality - review documentation content and structure
  • Provide final rating and recommendation

Phase 1: Root Cause Analysis ✅

Problem Being Solved: This PR addresses a process gap where the dogfood report from v2.2.3-dev testing was left as an untracked file after the dogfooding session completed. The dogfood skill was missing explicit instructions for committing the report to the repository.

Why This Approach:

  • Adds the completed dogfood report to version control for historical tracking
  • Updates the dogfood skill with "Phase 10 — Commit the Report" to prevent this gap in future sessions
  • Follows the skill's own improvement directive: "If during the dogfooding session you discover that this skill is missing steps... you are encouraged to edit it"

Risk Assessment: Very low risk - this is pure documentation with no code changes, configuration modifications, or functional impacts.

Phase 2: Configuration Change Detection ✅

No high-risk changes detected:

  • No biome/linting rule modifications
  • No test configuration changes
  • No build or module resolution modifications
  • No TypeScript configuration changes
  • Only documentation files modified (.claude/skills/dogfood/SKILL.md and new report file)

Phase 3: Test Integrity ✅

No test-related red flags:

  • No test files modified
  • No test assertions weakened or removed
  • No test timeouts or coverage thresholds lowered
  • This is pure documentation addition with no impact on test suite

Phase 4: Code Quality ✅

Documentation Content Quality:

  • Dogfood report (generated/DOGFOOD_REPORT_v2.2.3-dev.44e8146.md): Comprehensive 372-line testing report following established format. Covers installation, cold start testing, full command sweep, rebuilds, engine comparison, and identified 2 bugs with fixes.
  • Skill update (.claude/skills/dogfood/SKILL.md): Adds clear procedural Phase 10 with branching logic for different scenarios (bug-fix PRs vs clean runs). Includes verification step.

Structure and Clarity:

  • Both additions follow existing patterns in the repository
  • The skill update is properly sequenced as Phase 10 after Phase 9 (Report writing)
  • Instructions are clear with concrete bash commands for both scenarios
  • Maintains consistency with existing skill phases

Root Cause Analysis

Problem Being Solved: Process improvement — ensuring dogfood reports are committed rather than left untracked
Why This Approach: Direct fix for identified gap, follows repository's self-improvement philosophy
Risk Assessment: Minimal — documentation-only changes with no functional impact

Critical Concerns

None identified. This is straightforward documentation maintenance.

Final Recommendation

  • Rating: ⭐⭐⭐⭐☆ (4/5)
  • Action: APPROVE
  • Reasoning: Clean documentation addition that solves a real process gap. High quality report content and sensible skill improvement. No configuration risks or test impacts. Only missing the 5th star due to being purely administrative rather than feature work.

@greptile-apps
Copy link
Copy Markdown
Contributor

greptile-apps bot commented Feb 24, 2026

Greptile Summary

This PR adds the dogfood report for v2.2.3-dev.44e8146 that was previously left as an untracked file, and updates the dogfood skill with a new Phase 10 to ensure future reports are always committed.

Key changes:

  • Added comprehensive dogfood report (generated/DOGFOOD_REPORT_v2.2.3-dev.44e8146.md) documenting the testing session that found 2 bugs (both fixed in PR fix(cli): graceful error for cycles, export, embed when no graph.db exists #79)
  • Updated .claude/skills/dogfood/SKILL.md with Phase 10 instructions for committing reports, covering both scenarios: adding to first bug-fix PR or creating dedicated PR for green-path sessions
  • Report is well-structured with 12 sections covering setup, cold start testing, command sweep, rebuild/staleness, engine comparison, release-specific tests, bugs found, and overall 8/10 rating

Confidence Score: 5/5

  • This PR is safe to merge with no risk
  • This is a documentation-only PR that adds a previously generated dogfood report and updates the dogfood skill with clear instructions. No code changes, no functionality changes, zero risk of introducing bugs.
  • No files require special attention

Important Files Changed

Filename Overview
.claude/skills/dogfood/SKILL.md Added Phase 10 with clear instructions for committing dogfood reports to prevent leaving them as untracked files
generated/DOGFOOD_REPORT_v2.2.3-dev.44e8146.md Comprehensive dogfood report documenting v2.2.3-dev.44e8146 testing, found 2 bugs (both fixed in PR #79), rated 8/10

Last reviewed commit: 8247d5d

Copy link
Copy Markdown
Contributor

@greptile-apps greptile-apps bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

2 files reviewed, no comments

Edit Code Review Agent Settings | Greptile

@carlos-alm carlos-alm merged commit 4ad3fbb into main Feb 24, 2026
17 checks passed
@carlos-alm carlos-alm deleted the docs/dogfood-report-v2.2.3-dev branch February 24, 2026 04:26
carlos-alm added a commit that referenced this pull request Mar 16, 2026
…in backlog

These two items deliver the highest immediate impact on agent experience
and graph accuracy without requiring Rust porting or TypeScript migration.
They should be implemented before any Phase 4+ roadmap work.

- #83: hook-optimized `codegraph brief` enriches passively-injected context
- #71: basic type inference closes the biggest resolution gap for TS/Java
carlos-alm added a commit that referenced this pull request Mar 16, 2026
* docs: move 3.16 items to new Phase 5 (Runtime & Extensibility)

Expand the deferred Phase 3 items into a dedicated phase after
TypeScript Migration with detailed descriptions for each sub-item:
event-driven pipeline, unified engine strategy, subgraph export
filtering, transitive confidence, query caching, config profiles,
pagination standardization, and plugin system.

Renumber subsequent phases 5-9 → 6-10 with all cross-references
updated.

* refactor: migrate integration tests to InMemoryRepository

Add openRepo() utility that accepts an injected Repository instance or
falls back to SQLite, enabling tests to bypass the filesystem entirely.

- Add openRepo(dbPath, opts) to src/db/connection.js
- Make findMatchingNodes and buildDependencyGraph polymorphic (accept
  db or Repository via instanceof check)
- Refactor triageData, sequenceData, communitiesData to use openRepo
- Convert triage, sequence, communities test fixtures to createTestRepo()
  fluent builder (sequence dataflow tests stay on SQLite)
- Mark ROADMAP 3.13 InMemoryRepository migration item complete

Impact: 10 functions changed, 11 affected

* fix: use instanceof SqliteRepository guard and validate openRepo opts

Impact: 2 functions changed, 18 affected

* docs: add Phase 4 (Native Analysis Acceleration) to roadmap

Add new Phase 4 covering the port of JS-only build phases to Rust:
- 4.1-4.3: AST nodes, CFG, dataflow visitor ports (~587ms savings)
- 4.4: Batch SQLite inserts (~143ms)
- 4.5: Role classification & structure (~42ms)
- 4.6: Complete complexity pre-computation
- 4.7: Fix incremental rebuild data loss on native engine
- 4.8: Incremental rebuild performance (target sub-100ms)

Bump old Phases 4-10 to 5-11 with all cross-references updated.
Benchmark evidence shows ~50% of native build time is spent in
JS visitors that run identically on both engines.
carlos-alm added a commit that referenced this pull request Mar 16, 2026
* docs: promote #83 (brief command) and #71 (type inference) to Tier 0 in backlog

These two items deliver the highest immediate impact on agent experience
and graph accuracy without requiring Rust porting or TypeScript migration.
They should be implemented before any Phase 4+ roadmap work.

- #83: hook-optimized `codegraph brief` enriches passively-injected context
- #71: basic type inference closes the biggest resolution gap for TS/Java

* docs: add Phase 4 (Native Analysis Acceleration) to roadmap

Add new Phase 4 covering the port of JS-only build phases to Rust:
- 4.1-4.3: AST nodes, CFG, dataflow visitor ports (~587ms savings)
- 4.4: Batch SQLite inserts (~143ms)
- 4.5: Role classification & structure (~42ms)
- 4.6: Complete complexity pre-computation
- 4.7: Fix incremental rebuild data loss on native engine
- 4.8: Incremental rebuild performance (target sub-100ms)

Bump old Phases 4-10 to 5-11 with all cross-references updated.
Benchmark evidence shows ~50% of native build time is spent in
JS visitors that run identically on both engines.

* docs: fix sub-section numbering to match parent phase headings
carlos-alm added a commit that referenced this pull request Mar 17, 2026
* docs: promote #83 (brief command) and #71 (type inference) to Tier 0 in backlog

These two items deliver the highest immediate impact on agent experience
and graph accuracy without requiring Rust porting or TypeScript migration.
They should be implemented before any Phase 4+ roadmap work.

- #83: hook-optimized `codegraph brief` enriches passively-injected context
- #71: basic type inference closes the biggest resolution gap for TS/Java

* fix: include constant nodes in edge building and downstream queries (#487)

The edge-building stage excluded 'constant'-kind nodes from the node
lookup maps, so no import/dependency edges were ever created for
exported constants. This made them invisible to where, fn-impact,
query, roles, and all other graph queries. Also adds missing 'record'
kind to build-edges.js to match CORE_SYMBOL_KINDS.

Impact: 6 functions changed, 33 affected

* fix: include constant kind in whereSymbol and findMatchingNodes queries (#487)

whereSymbolImpl used ALL_SYMBOL_KINDS (core 10, no constant) so
codegraph where could not find constants. findMatchingNodes defaulted
to FUNCTION_KINDS which also excluded constant. Switch whereSymbolImpl
to EVERY_SYMBOL_KIND and add constant to FUNCTION_KINDS default.

Impact: 1 functions changed, 6 affected
carlos-alm added a commit that referenced this pull request Mar 17, 2026
* docs: promote #83 (brief command) and #71 (type inference) to Tier 0 in backlog

These two items deliver the highest immediate impact on agent experience
and graph accuracy without requiring Rust porting or TypeScript migration.
They should be implemented before any Phase 4+ roadmap work.

- #83: hook-optimized `codegraph brief` enriches passively-injected context
- #71: basic type inference closes the biggest resolution gap for TS/Java

* docs: add Phase 4 (Native Analysis Acceleration) to roadmap

Add new Phase 4 covering the port of JS-only build phases to Rust:
- 4.1-4.3: AST nodes, CFG, dataflow visitor ports (~587ms savings)
- 4.4: Batch SQLite inserts (~143ms)
- 4.5: Role classification & structure (~42ms)
- 4.6: Complete complexity pre-computation
- 4.7: Fix incremental rebuild data loss on native engine
- 4.8: Incremental rebuild performance (target sub-100ms)

Bump old Phases 4-10 to 5-11 with all cross-references updated.
Benchmark evidence shows ~50% of native build time is spent in
JS visitors that run identically on both engines.

* docs: fix sub-section numbering to match parent phase headings

* fix: align version computation between publish.yml and bench-version.js

- Add COMMITS=0 guard in publish.yml to return clean version when HEAD
  is exactly at a tag (mirrors bench-version.js early return)
- Change bench-version.js to use PATCH+1-dev.COMMITS format instead of
  PATCH+COMMITS-dev.SHA (mirrors publish.yml's new scheme)
- Fix fallback in bench-version.js to use dev.1 matching publish.yml's
  no-tags COMMITS=1 default

Impact: 1 functions changed, 0 affected

* feat: auto-detect semver bump in /release skill when no version provided

The release skill now scans commit history using conventional commit
rules to determine major/minor/patch automatically. Explicit version
argument still works as before.

* fix: restore SHA-based uniqueness in bench-version fallback and add -dev.0 suffix

The no-tags fallback hardcoded dev.1, causing repeated benchmark runs
to silently overwrite each other. Restore short SHA for uniqueness.
Also add -dev.0 suffix when COMMITS=0 so dev builds at an exact tag
are never confused with stable releases.

Impact: 1 functions changed, 0 affected

* fix: add -dev.0 suffix for COMMITS=0 dev builds in publish.yml

When HEAD is exactly at a release tag, the dev build emitted a clean
semver string identical to the stable release. Add -dev.0 suffix to
prevent confusion between dev pre-releases and stable releases.
carlos-alm added a commit that referenced this pull request Mar 17, 2026
* docs: promote #83 (brief command) and #71 (type inference) to Tier 0 in backlog

These two items deliver the highest immediate impact on agent experience
and graph accuracy without requiring Rust porting or TypeScript migration.
They should be implemented before any Phase 4+ roadmap work.

- #83: hook-optimized `codegraph brief` enriches passively-injected context
- #71: basic type inference closes the biggest resolution gap for TS/Java

* fix: include constant nodes in edge building and downstream queries (#487)

The edge-building stage excluded 'constant'-kind nodes from the node
lookup maps, so no import/dependency edges were ever created for
exported constants. This made them invisible to where, fn-impact,
query, roles, and all other graph queries. Also adds missing 'record'
kind to build-edges.js to match CORE_SYMBOL_KINDS.

Impact: 6 functions changed, 33 affected

* fix: include constant kind in whereSymbol and findMatchingNodes queries (#487)

whereSymbolImpl used ALL_SYMBOL_KINDS (core 10, no constant) so
codegraph where could not find constants. findMatchingNodes defaulted
to FUNCTION_KINDS which also excluded constant. Switch whereSymbolImpl
to EVERY_SYMBOL_KIND and add constant to FUNCTION_KINDS default.

Impact: 1 functions changed, 6 affected
carlos-alm added a commit that referenced this pull request Mar 17, 2026
* docs: promote #83 (brief command) and #71 (type inference) to Tier 0 in backlog

These two items deliver the highest immediate impact on agent experience
and graph accuracy without requiring Rust porting or TypeScript migration.
They should be implemented before any Phase 4+ roadmap work.

- #83: hook-optimized `codegraph brief` enriches passively-injected context
- #71: basic type inference closes the biggest resolution gap for TS/Java

* docs: add Phase 4 (Native Analysis Acceleration) to roadmap

Add new Phase 4 covering the port of JS-only build phases to Rust:
- 4.1-4.3: AST nodes, CFG, dataflow visitor ports (~587ms savings)
- 4.4: Batch SQLite inserts (~143ms)
- 4.5: Role classification & structure (~42ms)
- 4.6: Complete complexity pre-computation
- 4.7: Fix incremental rebuild data loss on native engine
- 4.8: Incremental rebuild performance (target sub-100ms)

Bump old Phases 4-10 to 5-11 with all cross-references updated.
Benchmark evidence shows ~50% of native build time is spent in
JS visitors that run identically on both engines.

* docs: fix sub-section numbering to match parent phase headings

* fix: align version computation between publish.yml and bench-version.js

- Add COMMITS=0 guard in publish.yml to return clean version when HEAD
  is exactly at a tag (mirrors bench-version.js early return)
- Change bench-version.js to use PATCH+1-dev.COMMITS format instead of
  PATCH+COMMITS-dev.SHA (mirrors publish.yml's new scheme)
- Fix fallback in bench-version.js to use dev.1 matching publish.yml's
  no-tags COMMITS=1 default

Impact: 1 functions changed, 0 affected

* feat: auto-detect semver bump in /release skill when no version provided

The release skill now scans commit history using conventional commit
rules to determine major/minor/patch automatically. Explicit version
argument still works as before.

* fix: restore SHA-based uniqueness in bench-version fallback and add -dev.0 suffix

The no-tags fallback hardcoded dev.1, causing repeated benchmark runs
to silently overwrite each other. Restore short SHA for uniqueness.
Also add -dev.0 suffix when COMMITS=0 so dev builds at an exact tag
are never confused with stable releases.

Impact: 1 functions changed, 0 affected

* fix: add -dev.0 suffix for COMMITS=0 dev builds in publish.yml

When HEAD is exactly at a release tag, the dev build emitted a clean
semver string identical to the stable release. Add -dev.0 suffix to
prevent confusion between dev pre-releases and stable releases.
@carlos-alm carlos-alm mentioned this pull request Mar 17, 2026
3 tasks
carlos-alm added a commit that referenced this pull request Mar 24, 2026
…arch snapshots (#557)

* docs: promote #83 (brief command) and #71 (type inference) to Tier 0 in backlog

These two items deliver the highest immediate impact on agent experience
and graph accuracy without requiring Rust porting or TypeScript migration.
They should be implemented before any Phase 4+ roadmap work.

- #83: hook-optimized `codegraph brief` enriches passively-injected context
- #71: basic type inference closes the biggest resolution gap for TS/Java

* docs: add Phase 4 (Native Analysis Acceleration) to roadmap

Add new Phase 4 covering the port of JS-only build phases to Rust:
- 4.1-4.3: AST nodes, CFG, dataflow visitor ports (~587ms savings)
- 4.4: Batch SQLite inserts (~143ms)
- 4.5: Role classification & structure (~42ms)
- 4.6: Complete complexity pre-computation
- 4.7: Fix incremental rebuild data loss on native engine
- 4.8: Incremental rebuild performance (target sub-100ms)

Bump old Phases 4-10 to 5-11 with all cross-references updated.
Benchmark evidence shows ~50% of native build time is spent in
JS visitors that run identically on both engines.

* docs: fix sub-section numbering to match parent phase headings

* fix: align version computation between publish.yml and bench-version.js

- Add COMMITS=0 guard in publish.yml to return clean version when HEAD
  is exactly at a tag (mirrors bench-version.js early return)
- Change bench-version.js to use PATCH+1-dev.COMMITS format instead of
  PATCH+COMMITS-dev.SHA (mirrors publish.yml's new scheme)
- Fix fallback in bench-version.js to use dev.1 matching publish.yml's
  no-tags COMMITS=1 default

Impact: 1 functions changed, 0 affected

* feat: auto-detect semver bump in /release skill when no version provided

The release skill now scans commit history using conventional commit
rules to determine major/minor/patch automatically. Explicit version
argument still works as before.

* feat: add /titan-run orchestrator with diff review, semantic assertions, and architectural snapshot

Add /titan-run skill that dispatches the full Titan pipeline (recon → gauntlet →
sync → forge) to sub-agents with fresh context windows, enabling end-to-end
autonomous execution.

Hardening layers added across the pipeline:
- Pre-Agent Gate (G1-G4): git health, worktree validity, state integrity, backups
- Post-phase validation (V1-V15): artifact structure, coverage, consistency checks
- Stall detection with per-phase thresholds and no-progress abort
- Mandatory human checkpoint before forge (unless --yes)

New validation tools integrated into forge and gate:
- Diff Review Agent (forge Step 9): verifies each diff matches the gauntlet
  recommendation and sync plan intent before gate runs
- Semantic Assertions (gate Step 5): export signature stability, import resolution
  integrity, dependency direction, re-export chain validation
- Architectural Snapshot Comparator (gate Step 5.5): community stability, cross-domain
  dependency direction, cohesion delta, drift detection vs pre-forge baseline

* fix: correct undefined variable in titan-run gauntlet efficiency check

* fix: address Greptile review feedback on titan-run skill

- Fix undefined previousAuditedCountBeforeAgent variable in gauntlet
  efficiency check (save pre-update count before reassignment)
- Add AU, UA, DU, UD to merge conflict detection markers
- Add warning when --start-from forge runs without arch-snapshot.json

* fix: address Greptile review feedback on titan-run skill

- Fix undefined previousAuditedCountBeforeAgent variable in gauntlet
  efficiency check (save pre-update count before reassignment)
- Add AU, UA, DU, UD to merge conflict detection markers
- Add warning when --start-from forge runs without arch-snapshot.json

* fix: restore FORGE to codegraph exports and fn-impact in tool table

Both commands are called in titan-forge's new diff review step (Step 9):
fn-impact for deletion audit (D4) and exports for re-export chain checks.

* fix: unstage files before restoring working tree in forge Step 13 rollback

* fix: add GATE to exports/fn-impact and FORGE to context in command table

* fix: update --yes description to reflect actual orchestrator scope

* chore: checkpoint stale working tree changes from prior sessions

Impact: 43 functions changed, 46 affected

* Revert "chore: checkpoint stale working tree changes from prior sessions"

This reverts commit a296b58.

* fix: correct command table and --yes flag documentation

* fix: address titan-gate review feedback — Louvain drift, temp paths, rollback range, FAIL template

- Use codegraph communities --drift for A1 comparison instead of raw IDs
- Add unique /tmp paths per invocation (epoch+PID) with cleanup
- Update auto-rollback exception from Steps 1-3,5-7 to Steps 1-3,6-8
- Split FAIL template: test/lint (rollback) vs structural/semantic (preserved)

* fix: address titan-run review feedback — --yes scope and --start-from validation

- Update --yes description to include forge per-phase confirmation
- Add Step 0.5 artifact pre-validation for --start-from skipped phases

* fix: titan-run review — Rules exceptions, skip-flag validation, test runner

* fix: persist arch temp dir path to file instead of shell variable (#557)

TITAN_TMP_ID was set in one Bash tool invocation but referenced in
later invocations. Shell variables don't survive between separate
Bash calls. Now uses mktemp -d and writes the path to
.codegraph/titan/.arch-tmpdir for cross-invocation recovery.

* fix: capture git SHA in shell before node -e to avoid unevaluated substitution (#557)

* fix: extend no-rollback exception to include semantic failures (Steps 5-8) (#557)

* fix: add explicit DIFF WARN verdicts to D5 leftover check (#557)

* fix: remove duplicate diff-impact call in Step 5c, add before/after comparison for Step 5d (#557)

* fix: correct stale step reference and use unique temp path in titan-gate (#557)

* fix: remove stale --yes from argument-hint and add D5 explicit verdict in titan-forge (#557)

* fix: address open review items in titan-gate and titan-forge (#557)

- Disambiguate no-rollback step range: "Steps 1-3, 5-8" → "Steps 1-3, 5, 5.5, 6-8"
  so AI agents include Step 5.5 (architectural snapshot) in the exception
- Guarantee Step 5.5 cleanup runs even on failure/early exit paths
- Fix Step 13 heading: exclude diff-review (handled by Step 9's own rollback)
- Replace hardcoded npm test in gate Step 4 with test runner detection
- Applied to both .claude/skills/ and docs/examples/ copies

* fix: address round 2 Greptile feedback on titan-gate and titan-forge (#557)

- Fix BARREL_TMP cleanup: use sidecar file pattern (.barrel-tmp) to
  persist temp path across shell invocations, matching the existing
  .arch-tmpdir pattern
- Remove redundant codegraph check in Step 5b: reuse results already
  collected in Step 2 instead of re-running the command
- Spell out D4 deletion audit: add explicit guidance on identifying
  deleted symbols via pre-change file comparison and git diff
- All changes applied to both .claude/skills/ and docs/examples/ copies

* fix: correct Step 5b reference from Step 2 to Step 1 in titan-gate (#557)

* fix: replace bash process substitution with temp file in D4 deletion audit (#557)

* fix: clarify A2 boundary check uses diff-impact edges not snapshot (#557)

* fix: restore FORGE to complexity --health command table row (#557)

* fix: add V3 snapshot list to orchestrator Rules exception list (#557)

* fix: replace A4 new-drift-warning check with resolved-drift positive signal (#557)

A4 overlapped with A1 producing false positives for touched symbols and
duplicate warnings for untouched ones. A1 already covers new drift
detection with severity escalation. A4 now only reports resolved drift
as a positive architectural improvement note.

* fix: add RUN to communities/structure rows and snapshot list entry in command table (#557)

titan-run Step 3.5a directly calls codegraph communities, structure,
and snapshot list. The README command table was missing these entries.

* fix(skill): distinguish semantic vs test/lint gate failures in forge rollback (#557)

* fix(skill): add V3 to pre-validation, fix efficiency/stall overlap (#557)

* fix(skill): add FAIL message template to Step 5d barrel export check (#557)

* fix(titan-run): track per-target progress in forge stall detection (#557)

Forge stall detection only compared completed phases, causing false
aborts when a multi-target phase required several sub-agent iterations.
Now also tracks completedTargets count — stallCount only increments
when neither phases nor targets advance.

* fix: remove duplicate previousCompletedPhases assignment in titan-run (#557)

* fix: scope A2 domain check to new edges only in titan-gate (#557)

* fix: guarantee Step 5d barrel cleanup runs on FAIL verdict (#557)

* fix(titan-gate): use Step 1 --staged results for boundary check (#557)

Step 5c referenced codegraph check --boundaries without --staged,
causing false positives from pre-existing violations. Now references
the already-collected Step 1 results instead.

* fix(titan-forge): clarify rollback, document --yes, init diffWarnings (#557)

- Step 11 semantic FAIL: clarify forge does its own rollback (not gate)
- Re-add --yes to argument list as documented passthrough
- Add diffWarnings: [] to execution state init block
- Document pre-gate test as fast-fail optimization with tradeoff note

* fix(titan-gate): replace codegraph deps with path-based layer inference in A2

codegraph deps is a file-level dependency map, not a symbol lookup.
The diff-impact edge output already contains source/target file paths,
so the domain/layer can be inferred directly from the path using the
GLOBAL_ARCH.md domain map without an extra command.

* fix(titan-forge): handle dead-code targets in D2 intent-match check

Dead-code targets have no gauntlet.ndjson entry, causing D2 to fail
silently. Now checks titan-state.json deadSymbols first and validates
the diff shows only deletions without needing a gauntlet lookup.

* fix(titan-run): add try/catch to arch-snapshot builder script

The node -e script that assembles arch-snapshot.json had no error
handling — if any input file was missing or malformed, it failed
silently and no snapshot was written. Now catches errors, prints a
warning, and continues without the snapshot.

* fix(titan-run): document explicit G3 state recovery procedure

G3 referenced backup recovery without documenting the steps. Now
includes: check backup exists, validate it is valid JSON before
restoring, cp .bak over the corrupt file, and stop if backup is
also corrupt or missing.

* fix(titan-run): guard NDJSON integrity check against missing file

The node -e script crashed with ENOENT if gauntlet.ndjson did not
exist yet. Now checks fs.existsSync first and outputs a result with
missing:true so the caller can handle it gracefully.

* fix(docs): remove stale GATE consumer from codegraph deps command table

Gate A2 no longer calls codegraph deps — it uses path-based layer inference
from diff-impact edge output instead. The command table was never updated to
reflect this change.

* fix(titan-run): make checkpoint snapshot status conditional on capture success

* fix(titan-forge): add explicit DIFF FAIL rollback commands and diffWarnings schema

* fix(skill): set autoConfirm=true after checkpoint so forge sub-agents get --yes (#557)

* fix(skill): add --yes to titan-forge argument-hint frontmatter (#557)

The --yes flag was documented in the Arguments section but missing
from the frontmatter hint that Claude Code displays in autocomplete.

* fix(skill): remove D1 from diffWarnings schema — D1 only produces FAIL (#557)

* fix(skill): clarify unexpected-commits as context-gathering in titan-run (#557)

* fix(skill): document A3/A4 committed-state limitation in titan-gate (#557)

* fix: handle arch capture failures and skip V13 when no commits (#557)

* fix(skill): preserve file extension in temp files for D4 and barrel validation (#557)

mktemp without an extension creates files that codegraph cannot
parse — LANGUAGE_REGISTRY matches by extension, so extension-less
temp files silently return no symbols, making D4 deletion audit
and Step 5d barrel validation no-ops.

* fix(skill): surface diffWarnings count in titan-run final report (#557)

* fix(skill): use diff-impact edges for titan-gate Step 5b import check (#557)

* fix(skill): skip currentPhase check in Step 0.5 pre-validation (#557)

* fix(skill): add RUN as reader for gauntlet, sync, and gate-log artifacts (#557)

* fix(skill): add codegraph where to README command table (#557)

* fix(skill): add step 3.5 skip condition, move A2 before snapshot gate (#557)

- Step 3.5 now has explicit skip condition for --start-from forge when arch-snapshot.json exists
- A2 dependency direction check moved before snapshot gate so it runs on standalone invocations
- Checkpoint template now uses explicit file-existence check for snapshot status

* fix(skill): init autoConfirm from --yes at parse time, fix snapshotStatus ordering (#557)

* fix(skill): add GLOBAL_ARCH.md guard to A2 dependency direction check (#557)

* fix(skill): clarify diffWarnings append, A3 sources, V13 test detection (#557)

* fix(skill): add explicit test runner detection to forge Step 10 (#557)

Forge Step 10 used a placeholder <detected-test-command> without actual
detection logic. This caused false-positive rollbacks on projects with
non-standard test scripts (e.g. test:ci but no test). Added the same
node -e detection block and NO_TEST_SCRIPT guard that V13 in titan-run
already uses. Applied to both .claude/skills/ and docs/examples/ copies.

* fix: increase vitest hookTimeout to 30s for Windows CI (#557)

The beforeAll hooks in batch.test.js and dataflow.test.js were timing
out at the default 10s on Windows Node 20 CI. The TS migration loader
hooks in NODE_OPTIONS add overhead that pushes SQLite initialization
past the default hookTimeout on slower CI runners. Align hookTimeout
with the existing testTimeout of 30s.

* fix(skill): move V13 failure-stop inside test block, add divergence warning (#557)

V13 test-failure condition was split outside the test-run block, causing
undefined testCmd reference for agents that don't short-circuit. Moved
the Stop path inside the same conditional as the test run.

Added divergence check at forge checkpoint (Step 3.5b) to warn when main
has advanced since the initial sync, preventing silent drift on long runs.

* fix(skill): cycle FAIL triggers rollback, Step 13 uses git diff for file discovery (#557)

Gate auto-rollback exception now correctly excludes Step 2 (cycle check)
from the no-rollback list. A new-cycle FAIL triggers the same rollback
path as lint/build/test failures, preventing graph DB inconsistency.

Forge Step 13 now discovers dirty files at rollback time using
git diff --cached/--name-only instead of relying on the agent carrying
the file list forward from Step 8, which can fail under context pressure.

* fix(skill): expand --start-from hint, add || true to divergence fetch (#557)

* fix: align .claude/** glob prefix with existing exclude patterns (#557)

* fix(skill): separate Step 5c and A2 boundary checks to avoid duplicate FAIL verdicts (#557)

* fix(skill): add cycle to forge Step 11 routing and gate rollback labels (#557)

* fix(skill): clarify FAIL labels, complexity routing, stale fetch (#557)

- Gate Step 9: replace misleading "advisory" label with
  explicit "structural and semantic failures" wording
- Forge Step 11: replace "semantic/structural" catch-all
  with explicit listing of all non-rollback failure types
- Run Step 3.5b: divergence check now detects fetch failure
  and skips instead of computing against stale origin/main

* fix(skill): move divergence check before checkpoint print block (#557)

* fix(skill): disambiguate D3|D5 schema notation in diffWarnings (#557)

* fix(skill): add new-file guard to Step 5a and explicit dedup tracking for Step 5c/A2 (#557)

* fix(skill): include failedTargets in forge stall detection to avoid false stalls (#557)

* fix(skill): detect yarn/pnpm/bun in test command detection via lock file (#557)

* fix(skill): replace <test-runner> placeholder with lock-file detection (#557)

* fix(skill): use dynamic file discovery in DIFF FAIL rollback instead of placeholder (#557)

* fix(skill): resolve cycle-rollback ambiguity across gate/forge/run (#557)

* fix(skill): detect yarn/pnpm/bun for lint and build commands in gate Step 4 (#557)

* fix(skill): update forge Rules rollback description to include unstage step (#557)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant