Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/auerbachb/claude-code-config/babysit-prnpx skills add auerbachb/claude-code-config --skill babysit-prgit clone --depth 1 https://github.com/auerbachb/claude-code-configWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/auerbachb/claude-code-config/babysit-pr)<a href="https://agentmods.dev/skills/auerbachb/claude-code-config/babysit-pr"><img src="https://agentmods.dev/badge/skills/auerbachb/claude-code-config/babysit-pr.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00172 | $0.16560 |
| Opus 5 | $0.00086 | $0.08280 |
| Sonnet 5 | $0.00034 | $0.03312 |
| Haiku 4.5 | $0.00017 | $0.01656 |
Grade A, and why
babysit-pr scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 862 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Watch one PR and drive it toward merge without looping forever. Each tick reads the PR's state through the shared scripts, classifies it, and dispatches the right skill — /fixpr to fix recoverable blockers, /wrap to merge when the gate is met — then re-arms the poll until a terminal condition fires.
/babysit-pr is a thin orchestrator: it never re-implements PR-state aggregation, the merge gate, fix logic, or the merge flow. It reads pr-state.sh + merge-gate.sh, and it dispatches /fixpr / /wrap. All code-verification, thread resolution, CI fixing, and merge-gate enforcement stay inside those skills (fixpr/SKILL.md, wrap/SKILL.md, cr-merge-gate.md). This skill only decides which to call and when to stop.
This is reused by /pr-monitor-and-manage (issue #460): that skill invokes /babysit-pr per discovered PR rather than re-implementing the per-PR decision tree.
Safety boundaries (HARD STOPS — non-negotiable, safety.md / #450)
/babysit-pr is read-only plus the dispatches below. It MUST NOT:
- Never babysit a PR you did not author (issue #733,
safety.md). Enrolling a PR in a watch loop is a "touch". Before the first tick, gate the target with"$PR_AUTHORSHIP_SH" <PR>(exit 0 = yours). Not yours / undetermined → refuse with one line naming the authorship guard, and do NOT enrol — unless the user named this specific PR in chat this session (per-PR override; say you are operating under it).polling-state-gate.sh --ensure-sessionalso refuses non-author enrolment as a fail-safe (bypass only with--allow-nonauthorunder that override). - Never modify branch protection — no calls to
.../branches/.../protection. - Never dismiss human-authored reviews. Only
/fixpr'sdismiss-stale-bot-changes.sh(bot allowlist, wrongcommit_id) may dismiss, and only bot reviews. - Never resolve a review thread itself — thread resolution happens only inside
/fixprSteps 1–4 after code-verification./babysit-prdoes not callresolveReviewThread. - Never bypass
/fixpr's code-verification step — it dispatches the full/fixprworkflow, never a shortcut. - Never post
@coderabbitai full reviewwithoutcr-review-hourly.sh --checkpassing first. The only sanctioned trigger path in this skill is the T1b pre-flight (pr-preflight.sh, issue #493), which gates CR oncr-review-hourly.sh(--check+ atomic--record-explicit) automatically, never triggers Greptile, and never flips another user's draft./fixprowns any further triggers after a push.
A human CHANGES_REQUESTED on HEAD is hard-blocked → record and exit. Never auto-dismiss it.
- Auto-resolve mode (
--auto-resolve-conflicts) performs unattended rebases and force-pushes. The resolver (.claude/skills/merge-conflict/resolve_merge_conflicts.py) only applies mechanically-simple hunks per its "when in doubt, complex" contract; any complex hunk aborts the rebase and reports the specific file + line range + reason instead of applying it. Because unattended force-pushes are a bigger authorization step than the rest of the dispatch table, this mode is opt-in — the default onCONFLICTINGis a conservative stop that recommends/merge-conflict.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 862 lines · 172 tokens per session scan A 663bb035c9e1
babysit-pr is a skill published in the GitHub repository auerbachb/claude-code-config (5 stars, last pushed 5d ago), licensed MIT. It adds 172 tokens to every session and 16,560 once invoked, about $0.0009 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
fixer
Surgical code fixer for Bug Hunter. Implements minimal, precise fixes for verified bugs. Uses doc-lookup (Context Hub + Context7) to verify correct API usage in patches. Respects fix strategy classifications (safe-autofix vs manual-review vs larger-refactor).
referee
Final arbiter for Bug Hunter. Receives Hunter findings and Skeptic challenges, independently re-reads code, and delivers authoritative verdicts with CVSS scoring and proof-of-concept generation for security findings.
do-competitively
Execute tasks through competitive multi-agent generation, meta-judge evaluation specification, multi-judge evaluation, and evidence-based synthesis.
adr
Create and manage Architecture Decision Records. Documents significant technical decisions with context, alternatives, and consequences. Also supports evaluate (assess proposals) and design (system design) modes.
deploy-checklist
Pre-deployment verification checklist. Generates customized checklists based on tech-stack with rollback triggers. Auto-invoked by /draft:upload.
judge-with-debate
Evaluate solutions through multi-round debate between independent judges until consensus.