* fix: preserve completed slice status on plan-milestone re-plan (#2558)
When plan-milestone re-plans a milestone that has already-completed slices,
the handler now checks existing slice status before inserting. Completed
slices retain their status instead of being reset to "pending".
Three changes:
1. handlePlanMilestone() checks getSlice() before insertSlice() and passes
the existing completed/done status instead of hardcoding "pending".
2. insertSlice() changed from INSERT OR IGNORE to INSERT ... ON CONFLICT
upsert that updates non-status fields (title, risk, depends, demo,
planning metadata) but preserves completed/done status at the DB layer.
3. reconcileWorktreeDb() slice and task merges now use LEFT JOIN to detect
existing completed rows in the main DB and never downgrade their status
when merging stale worktree data.
Closes#2558
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: relax completed-slice guard to allow re-plan when slices are retained
The #2960 guard blocked re-planning entirely when any completed slices
existed, conflicting with the #2558 preserve-completed-status logic.
Now only blocks when the new plan would drop completed slices.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix(gsd-db): prevent insertSlice ON CONFLICT from wiping populated fields
The ON CONFLICT clause unconditionally overwrote all non-status fields with
excluded values. Callers like complete-task.ts and complete-slice.ts use
insertSlice as an idempotent "ensure row exists" guard with only id and
milestoneId, causing defaults (empty strings, 0) to silently destroy
populated titles, demos, goals, and planning data.
Fix: use raw sentinel bind params (NULL when caller omitted the field) in
CASE guards so the ON CONFLICT UPDATE only overwrites fields the caller
actually provided. Initial INSERTs still get proper defaults.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* chore: retrigger CI
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: trek-e <trek-e@users.noreply.github.com>
* fix: reopen DB on cold resume and recognize U+2714 check mark
The paused-session resume path in auto.ts called rebuildState/deriveState
without first opening the project database, causing state derivation to
fall back to markdown parsing. This misparsed roadmap table rows with
glyph done markers and could redispatch wrong slices.
Export openProjectDbIfPresent from auto-start.ts and call it in the
resume path before rebuildState, matching the fresh bootstrap ordering.
Also add U+2714 (heavy check mark) to the table parser done-detection
regex alongside the existing U+2705/U+2611/U+2713 glyphs.
Closes#2940
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: remove duplicate openProjectDbIfPresent from rebase conflict
The rebase onto main introduced a duplicate `openProjectDbIfPresent`
function declaration (one from the PR, one from main), causing TS2393.
Keep the exported version that uses the static import.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: use logWarning instead of process.stderr in openProjectDbIfPresent
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: trek-e <trek-e@users.noreply.github.com>
* fix: dashboard model label shows dispatched model, not stale previous unit model
Move updateProgressWidget and ensurePreconditions after selectAndApplyModel
in phases.ts so the widget's first render tick reads the correct model.
Store currentDispatchedModelId in session state after model selection + hook
overrides, expose it via widgetStateAccessors, and update auto-dashboard.ts
to prefer the dispatched model ID over cmdCtx.model which can be stale.
Closes#2899
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix(test): widen runUnitPhase slice in ordering test to accommodate grown function body
The structural test for selectAndApplyModel ordering sliced only 8000
chars of runUnitPhase, but the function grew past that limit after
rebase, causing updateProgressWidget to fall outside the window.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* chore: retrigger CI
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: trek-e <trek-e@users.noreply.github.com>
* docs: add provider setup guide and improve onboarding hints
Fixes#2161
Add docs/providers.md with step-by-step setup instructions for every
supported LLM provider: OpenRouter, Ollama, LM Studio, vLLM, SGLang,
and all built-in providers. Includes env var names, example configs,
common pitfalls, and verification steps.
Improve onboarding wizard:
- Add URL hints to provider selection list
- Show common local endpoints when choosing Custom (OpenAI-compatible)
- Add post-setup guidance for OpenRouter and custom endpoints
- Reference docs/providers.md for compat troubleshooting
Update cross-references in getting-started.md, troubleshooting.md,
docs/README.md, and help-text.ts to link to the new guide.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* test: verify config help mentions OpenRouter, Ollama, and docs/providers.md
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: trek-e <trek-e@users.noreply.github.com>
The per-turn dedup cache introduced in the parent commit persists across
test cases since they all use the same question signature. Test 1 populates
the cache, causing tests 2 and 3 to get cached results instead of exercising
their intended code paths.
syncProjectRootToWorktree deleted empty gsd.db but left companion
-wal and -shm files on disk. On Node 24, node:sqlite attempts WAL
recovery from orphaned files, triggering a synchronous CPU spin loop
(227% CPU, 1.4GB RSS). Now deletes gsd.db-wal and gsd.db-shm when
the main DB is deleted or already missing.
Gate triggerTurn behind getInFlightToolCount() === 0 for both soft
timeout and context-pressure wrapup messages. Add clearQueue() to
stopAuto() and pauseAuto() to flush late async follow-ups.
Machine-local indexing state (LanceDB, sync cursors, job files) was
being tracked in Git, causing merge conflicts and stale cursor
propagation across branches. Gitignore alone doesn't affect
already-tracked files, so this removes them from the index while
keeping them on disk.
resolveDefaultSessionModel() previously only returned a result for
provider/model format strings, silently ignoring valid bare model IDs
like "gpt-5.4". This meant preferences could fail to override stale
settings.json defaults when users configured models without explicit
provider prefixes.
Now accepts sessionProvider param (ctx.model?.provider) to resolve bare
IDs. Also handles object configs without explicit provider field.
nextDecisionId() and nextRequirementId() compute the next ID via
SELECT MAX then pass it to a separate upsertDecision/upsertRequirement
call. When parallel tool calls hit these functions concurrently, both
read the same MAX value and produce the same ID — the second insert
silently overwrites the first.
Move the SELECT MAX + INSERT into a single transaction() call from
gsd-db.ts, which uses BEGIN/COMMIT/ROLLBACK and works on both
better-sqlite3 and node:sqlite providers. The transaction is
re-entrant safe (nested calls skip the BEGIN).
Same fix applied to saveRequirementToDb for consistency.
Closes#3326, closes#3339, closes#3459
Session bootstrap used ctx.model (from settings.json defaultProvider/defaultModel)
as the autoModeStartModel snapshot. When settings.json had a stale provider
(e.g. claude-code) but PREFERENCES.md was fully configured for openai-codex,
sessions would start with the wrong provider and fail with auth errors.
Add resolveDefaultSessionModel() to preferences-models.ts which extracts the
default model from GSD preferences (execution → planning → first configured).
In auto-start.ts, the preferred model now takes priority over ctx.model when
building startModelSnapshot, so PREFERENCES.md always wins over stale settings.
GSD-2 only searches ~/.agents/skills/ and .agents/skills/ for skills.
Claude Code's official skill directories (~/.claude/skills/ and
.claude/skills/) are not included in the search path, making GSD-2
blind to any skills managed there.
The skills.sh CLI (npx skills list -g) already recognises both
~/.agents/skills/ and ~/.claude/skills/ as valid global skill
directories. This commit aligns GSD-2's resolution logic with
that behaviour.
Affected functions:
- getSkillSearchDirs(): adds ~/.claude/skills/ and .claude/skills/
- captureAvailableSkills(): includes Claude Code dir in telemetry
- detectStaleSkills(): includes Claude Code dir in staleness checks
- detectNewSkills(): resolves SKILL.md from either directory
- isPackInstalled(): checks both dirs before recommending installs
- formatSkillDetail(): finds SKILL.md in either directory
The questionSignature() function only hashed sorted question IDs,
meaning calls with the same IDs but different text/options would
return stale cached answers. Now hashes the full canonicalized
payload (id, header, question, options, allowMultiple).
Adds 4 regression tests for signature correctness.
Add gsd_progress, gsd_roadmap, gsd_history, gsd_doctor, gsd_captures,
and gsd_knowledge tools that parse .gsd/ on disk — no session needed.
Inline lightweight readers in src/readers/ keep the package standalone
(zero new dependencies). 33 new tests, 64 total passing.
Address adversarial review findings:
1. [high] Override routing now requires an active auto-mode session
(in-process or remote via checkRemoteAutoSession) before writing
to a worktree path. Previously, any existing worktree directory
would receive the override even if no agent was running there —
a leftover worktree from a previous session would silently eat
the override.
2. [medium] Success messages now report the actual resolved override
location (worktree vs project root .gsd/OVERRIDES.md) so operators
know exactly where to look during recovery or manual rewrite.
Additional tests cover: inactive worktree fallback, double-gate
(autoRunning + valid .git), and getAutoWorktreePath null on missing .git.
Closes#3476
handleSteer used process.cwd() as the base path for appendOverride,
which writes to project/.gsd/OVERRIDES.md. When auto-mode runs in a
worktree, it reads from worktree/.gsd/ — so overrides written from a
second terminal were never seen by the agent.
Now checks for an active worktree via getAutoWorktreePath and writes
the override there when one exists, falling back to the project root
when no worktree is active.
Closes#3476
The codebase preferences block was accepted as a known key but never
validated or assigned in validatePreferences(), causing all user-configured
codebase defaults to be silently discarded. Adds validation for
exclude_patterns (string[]), max_files (positive int), and collapse_threshold
(positive int) with unknown-key warnings and 4 new tests.
Add configurable codebase map options via preferences.md (exclude_patterns,
max_files, collapse_threshold), expose --collapse-threshold as a CLI flag,
and auto-generate CODEBASE.md during project init for instant agent orientation.
Closes#3509
The ensureDbOpen catch block now logs via logWarning with error message
instead of structured diagnostic object. Update source-level assertion
to match the new pattern.
Update existing workflow-logger tests to use logError for audit
persistence assertions (warnings are now ephemeral). Add void
expression to empty catch blocks in detectMainBranch to satisfy
the no-empty-catch CI check.
Only persist error-severity entries to audit-log.jsonl (warnings stay
ephemeral in stderr + buffer). Sanitize persisted entries with message
truncation and context field allowlisting. Demote expected main/master
branch probe failures to silent control flow. Remove JSON.stringify of
diagnostic objects embedding cwd/paths in warning messages.
Addresses Codex adversarial review findings on workflow-logger migration.
workflow-events.ts: stop logging raw event line content to audit log —
log byte length only to avoid persisting potentially sensitive payload
fragments to .gsd/audit-log.jsonl.
parallel-orchestrator.ts: revert worker NDJSON parse failure to silent
drop — non-JSON lines (progress text, tool output) are expected in
worker stdout and logging each one creates I/O pressure and audit log
bloat in the parallel execution hot path.
Replace raw process.stderr.write(), console.error(), and empty catch
blocks across 50 GSD files with structured logWarning/logError calls
from the centralized workflow-logger system.
Add 13 new LogComponent types to cover all subsystems: recovery,
session, prompt, dashboard, timer, worktree, command, parallel, fs,
bootstrap, guided, registry, renderer.
Every migrated catch block now automatically:
- Shows in terminal (stderr) with component tag
- Gets buffered for auto-loop stuck-detection summary
- Persists to .gsd/audit-log.jsonl for post-mortem analysis
Update regression test to verify catch blocks use workflow-logger
instead of raw stderr/console, covering auto-mode files and all
explicitly migrated infrastructure files.
Closes#3506
Supersedes the approach in #3496
