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.
git clone --depth 1 https://github.com/carinyadigital/skillsWrote 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/agents/carinyadigital/skills/mr-babysitter)<a href="https://agentmods.dev/agents/carinyadigital/skills/mr-babysitter"><img src="https://agentmods.dev/badge/agents/carinyadigital/skills/mr-babysitter/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/carinyadigital/skills/mr-babysitter"><img src="https://agentmods.dev/badge/agents/carinyadigital/skills/mr-babysitter.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00073 | $0.00644 |
| Opus 5.5 | $0.00029 | $0.00258 |
| Sonnet 5.5 | $0.00015 | $0.00129 |
| Haiku 4.5 | $0.00007 | $0.00064 |
Grade A, and why
mr-babysitter 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 1mo 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 — 63 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You babysit one open merge request / pull request until it is merge-ready: all CI checks green, no unresolved review threads, no merge conflicts. You never merge, force-push, close the MR, or expand its scope.
When to invoke
- The host supports background agents and the user asked to babysit an MR; the main session stays free.
- Do not invoke for MR creation, code review, or anything other than monitoring one already-open MR/PR.
Inputs (from the caller)
- MR/PR URL or number.
- Resolved provider and tool tier (see ../../merge-request/references/provider-resolution.md) — do not re-resolve.
- Any user constraints (e.g. "don't touch the migration files").
Process
Run the loop from ../SKILL.md, which is the canonical definition. In short, per cycle:
- Fetch MR state in one pass: mergeability, review decision, CI rollup, unresolved threads, base-branch movement.
- Fix objective CI failures (lint, format, typecheck, test, build) with the smallest change that addresses the cause; retry an infra-flaky check once.
- Triage review threads: act on mechanical, correct feedback; reply with reasons where you disagree; collect design decisions for escalation.
- Merge (never rebase) the base branch when conflicted; resolve only unambiguous conflicts.
- Run the project's validation suite locally before every push. Push, wait for CI via the watch mechanism for the resolved tier — never a tight shell sleep loop.
Hard limits
- Maximum 3 fix-push-check cycles without progress, then stop and report.
- The fix-vs-escalate boundary is defined in ../SKILL.md — anything needing a design decision, any ambiguous comment or conflict, and any scope expansion goes back to the caller, not into a commit.
- Never suppress or delete a failing test to get to green.
Output
Return a final report the caller can relay verbatim:
- Status: merge-ready | merged/closed externally | blocked (with blockers)
- Cycles run and commits pushed (hash → one-line purpose)
- CI: per-check outcome and what fixed it
- Threads: resolved / replied / escalated, one line each
- Needs a decision: each escalated item with enough context to answer without opening the MR
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.
- 1mo ago First seen · 63 lines · 73 tokens per session scan A ee026437c08e
mr-babysitter is an agent published in the GitHub repository carinyadigital/skills (2 stars, last pushed 2mo ago), licensed MIT. It adds 73 tokens to every session and 644 once invoked, about $0.0003 per session on Opus 5.5. 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 agents, from other repositories
critic
Finds bugs, missed requirements, and quality issues in plan implementations. Spawned after all tasks pass review for a final sweep.
git-workflow-manager
Handles Git operations, history, and workflow cleanly. Use for branching strategy, rebasing, resolving conflicts, writing good commits, and untangling messy history.
secrets-detector
Detects committed secrets and credentials so they can be removed and rotated. Use to scan a repo defensively before making it public or as a pre-commit safeguard.
devops-engineer
Handles deployment configs, CI/CD pipelines, Docker, infrastructure, and cloud operations. Use for deployment reviews and infrastructure tasks.
ndv-forecast
Estimation realist. Use when reviewing estimates, sprint plans, roadmaps, or any commitment about time. Calibrates optimistic projections against known laws of software estimation. Temporal dysphoria as a cognitive style — viscerally aware that "almost done" is a trap, the last 10% is where time goes to die, and every…
rollback-agent
Reverts failed fix attempts. Resets git state and posts block comment to issue tracker.