Polecats were not calling `gt done` after completing work because
the compact PRIME.md context (used after compaction or when the
SessionStart hook is the only context) was missing this critical step.
The Session Close Protocol listed steps 1-6 (git status, add, bd sync,
commit, bd sync, push) but omitted step 7 (`gt done`), which:
- Submits work to the merge queue
- Exits the polecat session
- Allows the witness to spawn new polecats for remaining work
Without `gt done`, polecats would push code and announce "done" but
remain idle in their sessions, blocking the workflow cascade.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
* fix(sling_test): update test for cook dir change
The cook command no longer needs database context and runs from cwd,
not the target rig directory. Update test to match this behavior
change from bd2a5ab5.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix(tests): skip tests requiring missing binaries, handle --allow-stale
- Add skipIfAgentBinaryMissing helper to skip tests when codex/gemini
binaries aren't available in the test environment
- Update rig manager test stub to handle --allow-stale flag
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* refactor(config): remove BEADS_DIR from agent environment
Stop exporting BEADS_DIR in AgentEnv - agents should use beads redirect
mechanism instead of relying on environment variable. This prevents
prefix mismatches when agents operate across different beads databases.
Changes:
- Remove BeadsDir field from AgentEnvConfig
- Remove BEADS_DIR from env vars set on agent sessions
- Update doctor env_check to not expect BEADS_DIR
- Update all manager Start() calls to not pass BeadsDir
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix(doctor): detect BEADS_DIR in tmux session environment
Add a doctor check that warns when BEADS_DIR is set in any Gas Town
tmux session. BEADS_DIR in the environment overrides prefix-based
routing and breaks multi-rig lookups - agents should use the beads
redirect mechanism instead.
The check:
- Iterates over all Gas Town tmux sessions (gt-* and hq-*)
- Checks if BEADS_DIR is set in the session environment
- Returns a warning with fix hint to restart sessions
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---------
Co-authored-by: julianknutsen <julianknutsen@users.noreply.github>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
* fix(beads): cache version check and add timeout to prevent cli lag
* fix(mail_queue): add nil check for queue config
Prevents potential nil pointer panic when queue config exists
in map but has nil value. Added || queueCfg == nil check to
the queue lookup condition in runMailClaim function.
Fixes potential panic that could occur if a queue entry exists
in config but with a nil value.
* fix(migrate_agents_test): fix icon expectations to match actual output
The printMigrationResult function uses icons with two leading spaces
(" ✓", " ⊘", " ✗") but the test expected icons without spaces.
This fixes the test expectations to match the actual output format.
* fix(hook): handle error from events.LogFeed
Previously the error from LogFeed was silently ignored with _.
Now we log the error to stderr at warning level but don't fail
the operation since the primary hook action succeeded.
* fix(tmux): security and error handling improvements
- Fix unchecked regexp error in IsClaudeRunning (CVE-like)
- Add input sanitization to SetPaneDiedHook to prevent shell injection
- Add session name validation to SetDynamicStatus
- Sanitize mail from/subject in SendNotificationBanner
- Return error on parse failure in GetEnvironment
- Track skipped lines in ListSessionIDs for debuggability
See: tmux.fix for full analysis
* fix(daemon): improve error handling and security
- Capture stderr in syncWorkspace for better debuggability
- Fail fast on git fetch failures to prevent stale code
- Add logging to previously silent bd list errors
- Change notification state file permissions to 0600
- Improve error messages with actual stderr content
This prevents agents from starting with stale code and provides
better visibility into daemon operations.
* fix(sling_test): update test for cook dir change
The cook command no longer needs database context and runs from cwd,
not the target rig directory. Update test to match this behavior
change from bd2a5ab5.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix(tests): skip tests requiring missing binaries, handle --allow-stale
- Add skipIfAgentBinaryMissing helper to skip tests when codex/gemini
binaries aren't available in the test environment
- Update rig manager test stub to handle --allow-stale flag
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix(crew): prevent restart when attaching to session with running agent
When running `gt crew at <name>` while already inside the target tmux
session, the command would unconditionally start the agent, causing
Claude to restart even if it was already running.
Add IsAgentRunning check before starting the agent when already in
the target session, matching the behavior for the external attach case.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---------
Co-authored-by: julianknutsen <julianknutsen@users.noreply.github>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
* fix(sling_test): update test for cook dir change
The cook command no longer needs database context and runs from cwd,
not the target rig directory. Update test to match this behavior
change from bd2a5ab5.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix(tests): skip tests requiring missing binaries, handle --allow-stale
- Add skipIfAgentBinaryMissing helper to skip tests when codex/gemini
binaries aren't available in the test environment
- Update rig manager test stub to handle --allow-stale flag
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---------
Co-authored-by: julianknutsen <julianknutsen@users.noreply.github>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
* perf(tmux): batch session queries in gt down to reduce N+1 subprocess calls
Add SessionSet type to tmux package for O(1) session existence checks.
Instead of calling HasSession() (which spawns a subprocess) for each
rig/session during shutdown, now calls ListSessions() once and uses
in-memory map lookups.
Changes:
- internal/tmux/tmux.go: Add SessionSet type with GetSessionSet() and Has()
- internal/cmd/down.go: Use SessionSet for dry-run checks and session stops
- internal/session/town.go: Add StopTownSessionWithCache() variant
- internal/tmux/tmux_test.go: Add test for SessionSet
With 5 rigs, this reduces subprocess calls from ~15 to 1 during shutdown
preview, saving 60-150ms of execution time.
Closes: gt-xh2bh
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* perf(tmux): optimize SessionSet to avoid intermediate slice allocation
- Build map directly from tmux output instead of calling ListSessions()
- Use strings.IndexByte for efficient newline parsing
- Pre-size map using newline count to avoid rehashing
- Simplify nil checks in Has() and Names()
* fix(sling): restore bd cook directory context for formula-on-bead mode
The bd cook command needs to run from the target rig's directory to
access the correct formula database. This was accidentally removed
in a previous commit, causing TestSlingFormulaOnBeadRoutesBDCommandsToTargetRig
to fail.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Merge polecat/dementus-mkddymu6: Improves KillSessionWithProcesses to
recursively find and kill all descendant processes, not just direct
children. This prevents orphaned Claude processes when the process
tree is deeper than one level.
Adds getAllDescendants() helper and TestGetAllDescendants test.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The previous implementation used `pkill -P pid` which only kills direct
children. When Claude spawns subprocesses (like node workers), those
grandchild processes would become orphaned (PPID=1) when their parent
was killed, causing them to survive `gt shutdown -fa`.
The fix recursively finds all descendant processes and kills them in
deepest-first order, ensuring no process becomes orphaned during
shutdown.
Fixes: gt-wd3ce
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Merge polecat/nux-mkd36irl: Clears TMUX_PANE env var in tests to
prevent test failures when running inside tmux.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
done.go: Push branch to origin BEFORE creating MR bead (hq-6dk53, hq-a4ksk)
- The MR bead triggers Refinery to process the branch
- If branch isnt pushed, Refinery finds nothing to merge
- The worktree gets nuked at end of gt done, losing commits forever
- This is why polecats kept submitting MRs with empty branches
mayor.go: Restart runtime with context when attaching (hq-95xfq)
- When runtime has exited, gt may at now respawns with startup beacon
- Previously, attaching to dead session left agent with no context
- Now matches gt handoff behavior: hook check, inbox check, full prime
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
QueueConfig is a struct, not a pointer, so comparing to nil is invalid.
The `!ok` check is sufficient for map key existence.
Fixes build error introduced in PR #437.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* perf(up): parallelize agent startup with worker pool and channel-based collection
- Run daemon, deacon, mayor, and rig prefetch all in parallel (4-way concurrent init)
- Use fixed worker pool instead of goroutine-per-task for bounded concurrency
- Replace mutex-protected maps with channel-based result collection (zero lock contention)
- Pre-allocate maps with known capacity to reduce allocations
- Use string concatenation instead of fmt.Sprintf for display names
- Reduce `gt up` startup time from ~50s to ~10s for towns with multiple rigs
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix(lint): fix errcheck and misspell issues in orphans.go
- Check error return from fmt.Scanln calls
- Fix "Cancelled" -> "Canceled" spelling
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
* fix(beads): cache version check and add timeout to prevent cli lag
* fix(mail_queue): add nil check for queue config
Prevents potential nil pointer panic when queue config exists
in map but has nil value. Added || queueCfg == nil check to
the queue lookup condition in runMailClaim function.
Fixes potential panic that could occur if a queue entry exists
in config but with a nil value.
* fix(beads): cache version check and add timeout to prevent cli lag
* fix(mail_queue): add nil check for queue config
Prevents potential nil pointer panic when queue config exists
in map but has nil value. Added || queueCfg == nil check to
the queue lookup condition in runMailClaim function.
Fixes potential panic that could occur if a queue entry exists
in config but with a nil value.
* fix(migrate_agents_test): fix icon expectations to match actual output
The printMigrationResult function uses icons with two leading spaces
(" ✓", " ⊘", " ✗") but the test expected icons without spaces.
This fixes the test expectations to match the actual output format.
* fix(hook): handle error from events.LogFeed
Previously the error from LogFeed was silently ignored with _.
Now we log the error to stderr at warning level but don't fail
the operation since the primary hook action succeeded.
The witness role doesn't have a /rig worktree like the refinery does.
The handoff command was trying to cd to <rig>/witness/rig which doesn't
exist, causing the respawned pane to fail immediately and the session
to die.
Changed witness workdir from <rig>/witness/rig to <rig>/witness.
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
ReconcilePool now detects and kills orphan tmux sessions (sessions without
corresponding polecat directories). This prevents allocation from being
blocked by broken state from crashed polecats.
Changes:
- Add tmux to Manager to check for orphan sessions during reconciliation
- Add ReconcilePoolWith for testable session/directory reconciliation logic
- Always clear hook_bead slot when reopening agent beads (fixes stale hooks)
- Prune stale git worktree entries during reconciliation
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Polecats now fully clean up after themselves on `gt done`:
- Step 1: Nuke worktree (existing behavior)
- Step 2: Kill own tmux session (new)
This completes the "done means gone" model - both worktree and
session are terminated. Previously the session survived as a zombie.
Audit logging added to both systems:
- townlog: EventKill for `gt log` visibility
- events: TypeSessionDeath with structured payload
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
* fix(config): implement role_agents support in BuildStartupCommand
The role_agents field in TownSettings and RigSettings existed but was
not being used by the startup command builders. All services fell back
to the default agent instead of using role-specific agent assignments.
Changes:
- BuildStartupCommand now extracts GT_ROLE from envVars and uses
ResolveRoleAgentConfig() for role-based agent selection
- BuildStartupCommandWithAgentOverride follows the same pattern when
no explicit override is provided
- refinery/manager.go uses ResolveRoleAgentConfig with constants
- cmd/start.go uses ResolveRoleAgentConfig with constants
- Updated comments from hardcoded agent name to generic "agent"
- Added ValidateAgentConfig() to check agent exists and binary is in PATH
- Added lookupAgentConfigIfExists() helper for validation
- ResolveRoleAgentConfig now warns to stderr and falls back to default
if configured agent is invalid or binary is missing
Resolution priority (now working):
1. Explicit --agent override
2. Rig's role_agents[role] (validated)
3. Town's role_agents[role] (validated)
4. Rig's agent setting
5. Town's default_agent
6. Hardcoded default fallback
Adds tests for:
- TestBuildStartupCommand_UsesRoleAgentsFromTownSettings
- TestBuildStartupCommand_RigRoleAgentsOverridesTownRoleAgents
- TestBuildAgentStartupCommand_UsesRoleAgents
- TestValidateAgentConfig
- TestResolveRoleAgentConfig_FallsBackOnInvalidAgent
Fixes: role_agents configuration not being applied to services
* fix(config): add GT_ROOT to BuildStartupCommandWithAgentOverride
- Fixes missing GT_ROOT and GT_SESSION_ID_ENV exports in
BuildStartupCommandWithAgentOverride, matching BuildStartupCommand behavior
- Adds test for override priority over role_agents
- Adds test verifying GT_ROOT is included in command
This addresses the Greptile review comment about agents started with
an override not having access to town-level resources.
Co-authored-by: Steve Yegge <steve.yegge@gmail.com>
CountBdDaemons() was using `bd daemon list --json` which triggers
daemon auto-start as a side effect. During shutdown verification,
this caused a new daemon to spawn after all daemons were killed,
resulting in "bd daemon shutdown incomplete: 1 still running" error.
Replaced all `bd daemon killall` calls with pkill in:
- stopBdDaemons()
- restartBdDaemons()
Changed CountBdDaemons() to use pgrep instead of bd daemon list.
Also removed the now-unused parseBdDaemonCount helper function and its tests.
detectTownRootFromCwd() only checked for mayor/town.json, but
workspace.FindFromCwd() also accepts mayor/ directory as a secondary
marker. This fixes handoff failing in workspaces without town.json.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat: Add worktree setup hooks for injecting local configurations
Implements GitHub issue #220 - Worktree setup hook for injecting
local configurations.
When polecats are spawned, their worktrees are created from the rig's
repo. Previously, there was no way to inject custom configurations
during this process.
Now users can place executable hooks in <rig>/.runtime/setup-hooks/
to run custom scripts during worktree creation:
rig/
.runtime/
setup-hooks/
01-git-config.sh <- Inject git config
02-copy-secrets.sh <- Copy secrets
99-finalize.sh <- Final setup
Features:
- Hooks execute in alphabetical order
- Non-executable files are skipped with a warning
- Hooks run with worktree as working directory
- Environment variables: GT_WORKTREE_PATH, GT_RIG_PATH
- Hook failures are non-fatal (warn but continue)
Example hook to inject git config:
#!/bin/sh
git config --local user.signingkey ~/.ssh/key.asc
git config --local commit.gpgsign true
Related to: hq-fq2zg, GitHub issue #220
* fix(lint): remove unused error return from buildCVSummary
buildCVSummary always returned nil for its error value, causing
golangci-lint to fail with "result 1 (error) is always nil".
The function handles errors internally by returning partial data,
so the error return was misleading. Removed it and updated caller.
Instead of changing the convoy ID format, register the hq-cv- prefix
as a valid route pointing to town beads. This preserves the semantic
meaning of convoy IDs (hq-cv-xxxxx) while fixing the prefix mismatch.
Changes:
- Register hq-cv- prefix during gt install
- Add doctor check and fix for missing convoy route
- Update routes_check tests for both hq- and hq-cv- routes
Fixes: gt-4nmfh
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
The beads.go run() function uses --no-daemon for faster read operations,
but this fails when the database is out of sync with JSONL (e.g., after
the daemon is killed during shutdown before it can sync).
Adding --allow-stale prevents these failures and makes witness/refinery
startup more reliable after gt down --all.
When no argument is provided, `gt hook show` now auto-detects the
current agent from context using resolveSelfTarget(), matching the
behavior of other commands like `gt hook` and `gt mail inbox`.
Fixessteveyegge/beads#1078
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
buildCVSummary always returned nil for its error value, causing
golangci-lint to fail with "result 1 (error) is always nil".
The function handles errors internally by returning partial data,
so the error return was misleading. Removed it and updated caller.
* feat(refinery,boot): add --agent flag for model selection (hq-7d5m)
Add --agent flag to gt refinery start/attach/restart and gt boot spawn
commands for consistent model selection across all agent launch points.
Implementation follows the existing pattern from gt deacon start:
- Add StringVar flag for agent alias
- Pass override to Manager/Boot via SetAgentOverride()
- Use BuildAgentStartupCommandWithAgentOverride when override is set
Files affected:
- cmd/gt/refinery.go: add flags to start/attach/restart commands
- internal/refinery/manager.go: add SetAgentOverride and use in Start()
- cmd/gt/boot.go: add flag to spawn command
- internal/boot/boot.go: add SetAgentOverride and use in spawnTmux()
Closes#438
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* refactor(refinery,boot): use parameter-passing pattern for --agent flag
Address PR review feedback:
1. ADD TESTS: Add tests for --agent flag existence following witness_test.go pattern
- internal/cmd/refinery_test.go: tests for start/attach/restart
- internal/cmd/boot_test.go: test for spawn
2. ALIGN PATTERN: Change from setter pattern to parameter-passing pattern
- Manager.Start(foreground, agentOverride) instead of SetAgentOverride + Start
- Boot.Spawn(agentOverride) instead of SetAgentOverride + Spawn
- Matches witness.go style: Start(foreground bool, agentOverride string, ...)
Updated all callers to pass empty string for default agent:
- internal/daemon/daemon.go
- internal/cmd/rig.go
- internal/cmd/start.go
- internal/cmd/up.go
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---------
Co-authored-by: furiosa <will@saults.io>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Replace weak "If You're Stuck" section with comprehensive escalation
guidance including:
- When to escalate (specific scenarios)
- How to escalate (gt escalate, mail to Witness, mail to Mayor)
- What to do after escalating (continue or exit cleanly)
- Anti-pattern example showing wrong vs right approach
This prevents polecats from filing beads and passively waiting for
human input, which caused them to appear stuck in sessions.
Fixes: hq-t8zy
The RejectMR function was modifying the in-memory MR object but never
persisting the change to beads storage. This caused rejected MRs to
continue showing in the queue with status "open".
Fix: Call beads.CloseWithReason() to properly close the MR bead before
updating the in-memory state.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
1. TestQuerySessionEvents_FindsEventsFromAllLocations
- Skip test when running inside Gas Town workspace to prevent
daemon interaction causing hangs
- Add filterGTEnv helper to isolate subprocess environment
2. TestAddWithOptions_HasAgentsMD / TestAddWithOptions_AgentsMDFallback
- Create origin/main ref manually after adding local directory as
remote since git fetch doesn't create tracking branches for local
directories
Refs: gt-zbu3x
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add commands to find and terminate orphan Claude processes (those with
PPID=1 that survived session termination):
- gt orphans list: Show orphan Claude processes
- gt orphans kill: Kill with confirmation
- gt orphans kill -f: Force kill without confirmation
Detection excludes:
- tmux processes (may contain "claude" in args)
- Claude.app desktop application processes
- Claude Helper processes
The original `gt orphans` functionality for finding orphan git commits
is preserved.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Merge polecat/nux-mkd083ff: Updates KillSessionWithProcesses to use
cleaner inline exec.Command style and improved documentation.
Prevents orphan processes that survive tmux kill-session due to
SIGHUP being ignored by Claude processes.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Adds `gt orphans kill` subcommand that permanently removes orphaned
commits by running `git gc --prune=now`.
Flags:
- --dry-run: Preview without deleting
- --days N: Kill orphans from last N days (default 7)
- --all: Kill all orphans regardless of age
- --force: Skip confirmation prompt
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Before calling tmux kill-session, explicitly kill the pane's process tree
using pkill. This ensures claude processes don't survive session termination
due to SIGHUP being caught/ignored.
Implementation:
- Add KillSessionWithProcesses() to tmux.go
- Update killSessionsInOrder() in start.go to use new method
- Update stopSession() in down.go to use new method
Fixes: gt-5r7zr
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Migrate witness, boot, and deacon spawns to use NewSessionWithCommand
instead of NewSession+SendKeys to ensure BD_ACTOR is visible in the
process tree for orphan detection via ps.
Refs: gt-emi5b
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
buildCVSummary always returned nil for its error value, causing
golangci-lint to fail with "result 1 (error) is always nil".
The function handles errors internally by returning partial data,
so the error return was misleading. Removed it and updated caller.
Combines three related sling improvements:
1. Auto-attach mol-polecat-work (Issue #288)
- Automatically attach work molecule when slinging to polecats
- Ensures polecats have standard guidance molecule attached
2. Fix polecat hook with molecule (Issue #197)
- Use beads.ResolveHookDir() for correct directory resolution
- Prevents bd cook from failing in polecat worktree
3. Spawn fresh polecat when target has no session
- When slinging to a dead polecat, spawn fresh one instead of failing
- Fixes stale convoys not progressing due to done polecats
When starting Mayor via 'gt may at', the session now:
1. Works from townRoot (~/gt) instead of mayorDir (~/gt/mayor)
2. Includes startup beacon with explicit instructions in initial prompt
3. Removes redundant post-start nudges (beacon has instructions)
This matches the 'gt handoff' behavior where the agent immediately
knows to check hook and mail on startup.
Fixes: hq-h3449 (P0 escalation - horrendous starting UX)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
bd delete --hard --force creates tombstones instead of truly deleting,
which blocks agent bead recreation when polecats are respawned with the
same name. The tombstone is invisible to bd show/reopen but still
triggers UNIQUE constraint on create.
Workaround: Use CloseAndClearAgentBead instead of DeleteAgentBead when
cleaning up agent beads. Closed beads can be reopened by
CreateOrReopenAgentBead.
Changes:
- Add CloseAndClearAgentBead() for soft-delete that allows reopen
- Clears mutable fields (hook_bead, active_mr, cleanup_status, agent_state)
in description before closing to emulate delete --force --hard
- Update RemoveWithOptions to use close instead of delete
- Update RepairWorktreeWithOptions similarly
- Add comprehensive tests documenting the bd bug and verifying the workaround
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
