fix(polecat): refuse to stamp worktree-setup persistent branch on submit#8
Open
eric-jones wants to merge 1 commit into
Open
fix(polecat): refuse to stamp worktree-setup persistent branch on submit#8eric-jones wants to merge 1 commit into
eric-jones wants to merge 1 commit into
Conversation
eric-jones
force-pushed
the
topic/polecat-refuse-stamp-setup-branch
branch
from
4f21000 to
9fdfd1d
Compare
eric-jones
force-pushed
the
topic/polecat-refuse-stamp-setup-branch
branch
2 times, most recently
from
57c9847 to
6c5a2b0
Compare
The polecat prompt's FINAL REMINDER done sequence stamped `branch=$(git branch --show-current)` directly into the work bead's metadata. When the agent skipped the formula's workspace-setup step (which would create a fresh `polecat/<issue>` feature branch), `git branch --show-current` returned the per-slot persistent branch that `assets/scripts/worktree-setup.sh` creates — `gc-<agent>-<hash>`, where `<agent>` is the polecat slot name (e.g. `furiosa`) and `<hash>` is the worktree path's git hash-object value. That branch is shared across every polecat session that reuses the slot. So stamping it on a work bead points the refinery at a branch the next polecat in the slot will reuse for an unrelated bead — producing the failure where two work beads sit in the refinery queue pointing at the same branch, the refinery merges one (fast-forward), the other becomes stuck-conflicted, and no polecat ever picks it up because the slot's `git worktree list` view conflicts. ## What changed `examples/gastown/packs/gastown/agents/polecat/prompt.template.md` — FINAL REMINDER captures `CURRENT_BRANCH=$(git branch --show-current)`, runs a `case "$CURRENT_BRANCH" in gc-*) ... exit 1 ;;` guard, and stamps `--set-metadata branch="$CURRENT_BRANCH"` only after the guard passes. The error message names the formula step that creates the correct feature branch so the agent knows the next action. `examples/gastown/gastown_test.go` — new TestPolecatPromptRejectsWorktreeSetupBranchOnSubmit asserts the chained shape (capture, case-statement guard, exit 1, then stamp from the captured variable) and explicitly rejects the prior `--set-metadata branch=$(git branch --show-current)` shape so reverting the fix fails the test. ## Surface Pack-prompt edit only (no gascity Go code touched). The bug class is agent-behavior — instruction-shape per the maintenance-fixer solution-shape preference. The deeper architectural shape ( worktree-setup's persistent branch surviving across pool slot reuses) is left untouched: that branch is correct as a stable per-slot checkout target; the bug is the polecat prompt trusting `git branch --show-current` for bead metadata without validating it's a feature branch. ## Out of scope - Polecat formula `mol-polecat-work.toml` already creates the correct `polecat/<issue>` feature branch in workspace-setup; this PR does NOT modify that step. The guard catches the case where the agent goes straight to the prompt's done sequence without running workspace-setup. - gascity#1925 (open) adds the `git branch -D "$BRANCH"` cleanup that runs against the same `git branch --show-current` value — this guard layers cleanly on top: if `$BRANCH` matched `gc-*`, the guard exits before cleanup, surfacing the issue rather than silently deleting the persistent worktree branch. - The worktree-setup `branch_name()` function and its `gc-<agent>-<hash>` naming scheme are unchanged — that's the canonical worktree-branch contract; the fix just teaches the polecat to recognize it and refuse to stamp it as a feature branch. ## Testing - `go test ./examples/gastown/...` — full scoped suite passes (146s). - `TestPolecatPromptRejectsWorktreeSetupBranchOnSubmit` fails when the prompt is reverted (verified via stash + run). - Forensic evidence: `cities/validate-l5-20260510-v3/.gc/events.jsonl` shows two work beads (`vl2vr-73h`, `vl2vr-gcw`) both stamped with `branch=gc-furiosa-14808603229f` across two polecat sessions (`polecat-vl2v-e1b`, `polecat-vl2v-037k`) in the same `furiosa` slot. The hash `14808603229f` exactly matches `printf '%s' '<worktree-path>' | git hash-object --stdin | cut -c1-12` — i.e. `worktree-setup.sh`'s `branch_name()` output for that path. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> Upstream-PR: gastownhall#1959
eric-jones
force-pushed
the
topic/polecat-refuse-stamp-setup-branch
branch
from
6c5a2b0 to
c9d81af
Compare
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.
Fork-PR for review. Upstream: gastownhall#1959 WITHDRAWN 2026-05-20 (Q3 lifecycle-hook). Paired with topic/polecat-release-pool-slot. Retained on fork.