mr-babysit

An automated helper for merge requests (MRs), which are proposed code changes waiting for review. It repeatedly rebases them, fixes suitable CI failures, handles review comments, and waits for automated reviews.

In plain words
What is it for?
Use it to maintain one or more active merge requests until their checks pass and review activity is quiet. It evaluates review feedback before changing code.
Why use it?
It reduces the manual work of keeping merge requests ready to merge. It also separates routine fixes from product or design decisions that need the user.

Command

Part of the sdlc plugin — 1 skill, 7 commands shipped together

Install

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.

agentmods
npx agentmods add commands/fprochazka/claude-code-plugins/mr-babysit
Clone the repo
git clone --depth 1 https://github.com/fprochazka/claude-code-plugins

Or install sdlc, the plugin that ships this one along with the rest of its 1 skill, 7 commands.

Per session 30 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 8,559 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00030 $0.08559
Opus 5 $0.00015 $0.04279
Sonnet 5 $0.00006 $0.01712
Haiku 4.5 $0.00003 $0.00856

Measured 3d ago against content hash aef2ed3eb9b6, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

mr-babysit 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 3d 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.

plugins/sdlc/commands/mr-babysit.md · 285 lines

How it starts

The opening of the file, as written. The whole thing — 285 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Babysit the MR(s)

Run a continuous loop that watches one or more merge requests and drives each toward mergeable: keep it rebased, fix failing CI, resolve review comments that don't need product decisions, and wait for automated reviewers — passing back to the user only for genuine product/design calls or when everything solvable is solved.

This command is the author side of an MR. It is the mirror image of /code-review:watch, which is the reviewer side and never changes code.

This is not a rubber stamp. Review feedback — especially from AI review bots — is sometimes correct, sometimes wrong, sometimes pedantic, and sometimes proposes a fix that creates a new problem. Every finding is critically evaluated before any change lands.

Autonomy — you are pre-authorized; pausing to ask is a failure

This command intentionally runs at maximum autonomy, and for the duration of the loop that supersedes your usual defaults. Standing guidance you'd normally follow — global/project CLAUDE.md ("be restrained about implementing", "outline a plan and wait for my OK"), the general "confirm hard-to-reverse or outward-facing actions first" default — is deliberately overridden inside the babysit loop for the pre-authorized operations below. You already have the OK: the user invoked babysit precisely so these run unattended, and re-asking for it defeats the command. This override is scoped: it applies only to the named operations, only until the loop hands back, and it never extends to the explicit halt conditions under Stopping instead of handing over — a genuine product/design decision, an unresolvable conflict, or a cap hit still stop and ask. Normal caution resumes the moment the loop ends.

The loop's entire value is that it runs unattended. Within it you are explicitly authorized to git rebase, git push --force-with-lease, git push, retry CI jobs, and reply-to / resolve threads without stopping to ask — doing exactly these is the job, not a risk to escalate. Do not pause mid-loop for permission to force-push, rebase, or push; that defeats the purpose and is itself a failure of the loop. Stop and hand back only for the explicit halt conditions under Stopping instead of handing over — a genuine product/design decision, an unresolvable conflict, or a cap hit. Everything else, just do.

Read the full file on GitHub · 285 lines

Changes

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.

  1. 3d ago First seen · 285 lines · 30 tokens per session scan A aef2ed3eb9b6

Subscribe to this mod's changes

mr-babysit is a command published in the GitHub repository fprochazka/claude-code-plugins (11 stars, last pushed 5d ago), licensed MIT. It adds 30 tokens to every session and 8,559 once invoked, about $0.0002 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 2026-08-30.