Claude Skills by auerbachb

Merge a solo-owner PR whose only blocker is branch protection, after verifying the merge gate. Two shapes with different authority — Claude auto-runs the plain shape (a bare `gh pr merge --squash --admin`, no protection change); the `enforce_admins` toggle shape stays print-only for the repo admin. Triggers on /admin-merge and when /wrap or /babysit-pr hits a branch-protection block.

Cleanly stop an active /babysit-pr watcher for a PR. Sets the watcher's stop flag in session-state, stops its persistent Monitor task, and confirms. Invoke as `/babysit-pr-stop <PR>`.

Watch a single PR on a persistent Monitor and auto-dispatch /fixpr (recoverable blockers) or /wrap (merge-ready). Reads PR state via pr-state.sh + merge-gate.sh each tick, classifies into {merged, merge-ready, conflicting, has-recoverable-blockers, waiting-on-bots, hard-blocked}, tracks in-flight dispatches in session-state for idempotency, applies stable-state backoff, emits a timestamped heartbeat per tick, and hard-terminates on merge/hard-blocked/N blocker ticks/user stop. Stop with /baby...

Resume work stopped by /end; `/go-on` is the primary resume entry point and routes here. Explicitly clears the current session's background-launch gate, reports preserved stopped tasks, and resumes selected recoverable work without duplicating live tasks. The refill gate stays paused unless --resume-refill is supplied. Triggers on "end-resume", "resume from end", "continue ended work", "resume ended work".

Use when usage allowance is exhausted or running thin and this session must stop for hours or days while preserving resumable work. Blocks new launches, gives running work a bounded checkpoint window, stops every remaining owned background task, then writes a portable handoff. Triggers on "end", "clean end", "wind down", "portable handoff", "hand this to another agent", "I'm running low".

Bounded-convergence PR cleanup — audit every review thread + every CI check-run, fix all issues, push once per sweep, dismiss stale bot CHANGES_REQUESTED on old commits, resolve all threads via GraphQL, then wait (capped) for bot verdicts + CI on the new SHA, re-sweeping on new findings up to 5 iterations. Zero uncollapsed threads and zero failing checks when done.

Use when stopped work should be picked back up, whatever stopped it — `/pause`, `/end`, a token-exhaustion handoff, a session that died (crash, compaction, sign-out), or a stalled review/merge workflow. Universal resume — classifies the stoppage from the evidence. Invoke as `/go-on [--resume-refill] [--again]`.

Use when asking whether the harness now does natively what we automate by hand — "audit our rules/skills/hooks", "what's redundant", "does Claude Code do this already", or the monthly check. Verdicts every rule, skill, script, and hook against live harness behavior, then files issues. Advisory only — never edits.

Capture-only thread mode for drafting and opening well-structured GitHub issues. Puts the thread into issue-capture mode — the only job is creating, editing, and closing issues (no implementation, no worktrees, no /start-issue). Auto-opens issues (no approval gate) and reports back a concise summary plus the scope calls it made as decision points; creates canonical 6-section bodies with functional-first tone, runs dedup search, auto-applies validated labels, supports batch + cross-references,...

Use when you name the wall-clock time you have to stop — "I need to leave at 7 PM", "hard stop at 5:30", "I'm out at 6". Arms that time as the repo's planning deadline so dispatch declines pipelines that cannot finish before it, then proactively checks in at a lead time before it and winds the thread down through /pause so everything is merged or resumable by the time you go. Works in /pm threads and any thread running sub-agent pipelines. Triggers on "I need to leave at", "hard stop at", "le...

Audit the durable memory store (~/.claude/projects/*/memory/) and suggest prunes. Reports orphaned .md files, dangling index pointers, index size vs the context budget, and advisory superseded/duplicate entries — then, only on explicit confirmation, deletes the chosen files, rewrites MEMORY.md, and verifies integrity. Never auto-deletes. Sibling to /pm-clean. Triggers on "memory-clean", "clean memory", "prune memory", "stale memories", "orphaned memory files".

Classify merge/rebase conflicts against main, auto-resolve safe (simple) hunks and stage clean files, and report complex hunks for human judgment. Does not commit.

Squash merge the current PR. Verifies merge gate and acceptance criteria before merging.

Audit every open PR for engagement from all four AI reviewers (CodeRabbit, CodeAnt, BugBot, Graphite). Builds a ✅/❌ gap matrix from pr-state.sh (reviews + comments + check-runs on HEAD), detects blockers (draft PRs, CR rate-limit), and posts the missing trigger comments after confirmation. Read-only audit by default; never auto-flips drafts or auto-merges. Triggers on "/monitor", "check all PRs for reviewer engagement", "make sure all 4 reviewers ran", "did codeant/cursor/graphite review #N".

Resume companion to /pause; `/go-on` is the primary resume entry point and routes here. Explicitly clears the background-launch gate, reads parked and stopped-task recovery state, prints the current board, and re-arms selected work without duplicating live tasks. The refill gate is cleared only with --resume-refill. Triggers on "pause-resume", "resume from pause", "back from laptop", "restore parked work", "what did I park".

Use when you are closing your laptop and need every current-session background task stopped at a resumable boundary. Blocks new launches, uses a bounded runway to land safe work, hard-stops leftovers, and writes machine-readable resume state. Triggers on "pause", "laptop close", "heading out", "park the work", "shutting down".

Scan two independent fronts and recommend cleanup — open GitHub issues (solved-by-PR, inactive, superseded, duplicates) and the on-disk workspace (stale worktrees + local/remote branches, via the shared stale-cleanup.sh). Never auto-closes issues or deletes worktrees/branches without confirmation. Triggers on "pm-clean", "stale issues", "clean backlog", "close stale", "clean worktrees", "stale branches", "prune branches".

One-shot triage of your open PRs with no activity above a configurable age threshold — classifies each as close or merge, renders a Forgotten PRs block, and dispatches confirmed merges as phase-c-merger subagents. Invoked inline by /pm Step 1D; can also be run standalone. Triggers on "pm-forgotten-pr", "forgotten PRs", "idle PRs", "stale PRs".

Generate a PM handoff prompt for context-window turnover. Captures static config, live GitHub state, in-flight thread state, and memory summary into a self-contained prompt for a fresh PM thread. Bootstraps `.claude/pm-config.md` on first run. Triggers on "pm-handoff", "handoff", "context turnover", "new pm thread", "thread turnover".

Manage the OKRs section of `.claude/pm-config.md`. Show current objectives, set new ones, or get AI-suggested updates based on recent work. OKRs drive `/pm` ranking. Use this whenever you want to view, set, or update project OKRs, objectives, key results, or goals. Triggers on "pm okr", "set okrs", "show okrs", "suggest okrs", "project goals", "objectives".

Re-scan the current repo and update `.claude/pm-config.md` with fresh infrastructure and architecture detection while preserving user-edited sections (Role, OKRs, Team, Notes, Dependency Rules, Workflow Rules), then run a stale-cleanup pass to prune long-abandoned worktrees and branches. Use this after major milestones, new service integrations, or significant directory restructuring. Triggers on "pm update", "update pm config", "refresh pm config", "rescan repo".

Active PM orchestrator — manages issue pipeline, tracks coding threads, ranks the open backlog (OKR-aware) against a business goal, and suggests next work. Cold-starts from GitHub state, resumes from a /pm-handoff prompt, or runs continuously all day via `/pm day`. Triggers on "pm", "project manager", "orchestrate", "what should I work on", "rank issues", "work the backlog all day".

Clean-cancel companion to /pr-monitor-and-manage. Stops the PR-fleet Monitor tasks, clears monitoring state in session-state.json, and prints a final summary. Triggers on "/pr-monitor-and-manage-stop", "/pmm-stop", "stop monitoring PRs".

Resume companion to /pr-monitor-and-manage. Wakes a paused PR-fleet manager from its saved pause marker, stops any auto-wake Monitor, and re-arms the main Monitor. Also serves as the lightweight --auto-check target for that re-scan. Triggers on "/pr-monitor-and-manage-wake", "/pmm-wake", "wake PR monitor".

Thread-level PR fleet manager. Rediscovers your open PRs every tick, prints a status table on the first tick, after a resume, at pause/stop, and whenever a decision is needed or you ask (one-line heartbeat otherwise), and auto-dispatches the per-PR decision tree (rebase / parallel phase-a-fixer / sequential /wrap) until the fleet is clean, hard-blocked, or idle. Auto-pauses when idle; resume with /pr-monitor-and-manage-wake. Triggers on "/pr-monitor-and-manage", "/pmm", "manage PRs", "PR flee...

Executive PR review from a CTO/CPO/Chief Data Scientist lens. Analyzes multiple PRs in parallel for strategic fit, risk, issue alignment, and operational readiness. Use when reviewing PRs before merge, assessing merge readiness, or doing a leadership pass on open PRs.

Analyze GitHub issues to assess complexity, recommend a model tier, and generate tailored prompts with pre-extracted context. Use when starting work, planning sprints, right-sizing model choice, or analyzing issue batches. When called with no args in a PM thread, auto-detects suggested issues, partitioning inline subagent runs (the default) from the few too-big issues that need a separate thread.

Produce a concise, conversational functional summary of what a PR or issue actually changed or hopes to accomplish — nested bullets by default, optional Markdown table. Functional / feature lens, not an implementation walkthrough.

Use when a Phase B agent receives bot review findings — before implementing any suggestion, especially when a finding conflicts with sibling patterns in the codebase or contradicts another reviewer. Judgment layer only; the mechanical fix/reply/resolve loop lives in cr-github-review.md.

Use when asking whether the AI review tools still earn what we pay for them — "audit the review stack", "are we still paying for Greptile", "did a cap change", "re-check review tool costs", or the monthly check. Re-measures each tool's billed state, caps, throughput, and unique value, compares against the recorded baseline, and files drift issues. Advisory only — never edits rules.

Generate a daily standup summary of what was accomplished since the last standup, from a business logic perspective. Reads PR bodies to understand what changes actually enabled.

End-to-end issue-to-coding workflow. Accepts an issue number or description, handles CR plan polling, merges plans into issue body, creates worktree and branch, outputs ready-to-code summary. Use to start work on an issue without manually walking through the issue-planning flow.

Show a dashboard of all open PRs with review state, unresolved findings, and blockers.

Use when designing a parallel agent dispatch, deciding whether to parallelize work, or writing isolated subagent prompts. Teaches the prompt-crafting discipline for independent parallel agents in the Phase A/B/C model.

Run issues inline as subagents directly from a PM thread — any tier, except issues too big for a subagent. Assesses subagent fit, spawns Phase A/B/C agents, monitors progress, and reports merge readiness. Use to execute selected issues inline instead of in separate coding threads.

Use when starting several backlog issues at once and needing to know which can safely run in parallel — "wave", "next wave", "what can I run in parallel". Returns a dependency- and overlap-filtered independent set, capped at the pipeline ceiling, offered as click-to-launch chips. Never launches anything.

End-of-session command — verify no unresolved findings, squash merge, sync main, detect per-PR follow-ups, run a full-session loose-ends sweep, and extract lessons. Nearly silent by default — a clean merge emits one line (`merged PR #N`) plus any items needing your decision; pass `--verbose` (or just ask) for the merged + follow-ups summary and the full per-phase report. Accepts an optional PR reference (`/wrap <URL>`, `/wrap #N`, `/wrap N`); with no argument it infers the PR from the current...