docs(skills): fix broken board commands, exit propagation, worktree policy, and skill contradictions#1851
Conversation
…olicy, and skill contradictions Fixes #1819 Co-Authored-By: Claude dev-team-lead (Sonnet 4.6) <noreply@anthropic.com>
|
[security-engineer] Config-only review of PR #1851 ( Scope confirmed: 1. Worktree cleanup bash sequences — safeThe new sequence ( git status --porcelain # must be empty before proceeding
CURRENT_BRANCH=$(git branch --show-current)
WORKTREE_PATH=$(pwd)
BASE_REPO=$(git worktree list --porcelain | awk '/^worktree/{print $2; exit}')
cd "$BASE_REPO"
git worktree remove "$WORKTREE_PATH"
git branch -D "$CURRENT_BRANCH"
Minor (informational, non-blocking): the 2. Release step 6 — improvement, no bare push to
|
|
🎉 This PR is included in version 2.13.0-beta.20 🎉 The release is available on GitHub release Your semantic-release bot 📦🚀 |
Summary
gh project item-list --queryboard-status command (no such flag) indevelop,epic-start, andrelease— replaced with the proven idempotentgh project item-add+item-editpattern/exitpropagating intobatch-develop/releaseloops invoking/develop— the first item's/exitwas terminating the whole batch/release session; excluded from mid-loop invocations, deferred to a single cleanup at the true end of each flowdevelop/release/batch-developwith CLAUDE.md's flipped Session Isolation policy (PR docs: flip worktree policy — clean up worktrees when work is complete #1825):git worktree remove+git branch -D(force, since squash-merges fail-d's ancestry check) instead of the removed policy's "do not remove" stance or the undefined/exitshortcutfix-e2e's unsatisfiable branch-naming convention (no issue number ever exists for this skill) and its missing--base betaon PR creationepic-start's step 5→6 approval gap (posted the plan but never waited for user approval before proceeding); added a matching autonomy carve-out inepic-rundependabot's step 5c/5d contradiction on who commits/pushes, and stopdev-team-lead [MODE: commit]from creating a duplicate PR for an already-existing Dependabot PRreview-pr's verdict-vocabulary mismatch (dev-team-lead [MODE: review]was instructed to post agh pr reviewit has no access to post) and its required-checks list (was listing non-required jobs, omittingE2E Gates)releasestep 6's baregit pushdirectly tobeta(now routed through a proper branch+PR+CI+squash-merge cycle) and its circular/undefined epic-enriched PR body reference in step 2b (now defines the actual template)epic-close's stale step-count claim (13→8) and deadauto-fix.ymllint check (repointed tonpm run lintdirectly, since no CI workflow runs lint)batch-develop,dependabot,fix-e2e,mini-epic,review-pr) to CLAUDE.md's Orchestration Skills tabledevelop's duplicate copy (co-located*.test.ts/*.test.tsxcommits now require aqa-integration-testertrailer instead of incorrectly demandingbackend-developer/frontend-developer)batch-develop's work-queue file to/tmp/batch-queue.mdto stop colliding with/tmp/notes.md's release-feedback channelCo-Authored-By: Claude Opus 4.6trailers with the established🤖 Generated with Claude Codefooter for orchestrator-authored PR bodies (no canonical trailer exists for the orchestrator itself)Fixes #1819
Test plan
/exit, stale policy text, trailer duplication, step counts, dead workflow refs, etc.)npx prettier --checkclean on all touched filesCo-Authored-By: Claude dev-team-lead (Sonnet 4.6) noreply@anthropic.com