skills.rest
nsollazzo
nsollazzoCOMMUNITY
Explore all skills by this author →
Website𝕏@n_sollazzoGitHub

babysit

Drives a GitHub pull request to approved, CI-green, mergeable state through iterative review-response passes.

Updated Jun 15, 2026
Search Tags:#code review#pull request#github#gh cli#ci monitoring#review automation#approval loop
One-click install
npx skills add https://github.com/nsollazzo/ship-kit --skill babysit-nsollazzo
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: babysit
Source: https://github.com/nsollazzo/ship-kit/tree/main/.agents/skills/babysit
Command: npx skills add https://github.com/nsollazzo/ship-kit --skill babysit-nsollazzo

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Keeping a pull request moving after submission is tedious: CI checks fail, reviewers leave findings, threads go unanswered, and approvals go stale. This Skill automates the full loop of watching a GitHub PR, verifying every review finding against the actual code, fixing what is real, pushing back with evidence on what is not, and replying on every thread until the PR reaches a fresh APPROVED state with all CI green and mergeable status. ## Core Features & Use Cases - Goal-driven review loop: Runs repeated passes (observe, react, evaluate) until the PR has a fresh non-self APPROVED review, all CI checks passing, and mergeable status — without ever merging, since a human retains that step. - Adversarial finding verification: Each review finding is checked against the actual code at the cited file and line, optionally in parallel sub-agents, so real bugs get minimal conventional-commit fixes while false positives get evidence-based pushback replies. - Thread and gate management: Replies on every review comment with the fix and commit SHA, resolves threads via the GitHub GraphQL API, distinguishes bot-posted red statuses from real CI failures, and guards against dismissing stale approvals. - Use Case: After opening a PR, a reviewer bot posts five inline findings. The Skill verifies each against the code, fixes the two real ones with surgical commits, replies with evidence on the three false positives, pushes, waits for re-review, and loops until approval lands. ## Quick Start Ask the agent to babysit PR number 42 in the current repository and drive it to a green approved state.

Frequently Asked Questions about babysit

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automatically respond to GitHub PR review comments?▼

Use the gh CLI to fetch unresolved review threads via the GraphQL API, verify each finding against the cited code, then post replies with the fix and commit SHA using the pulls comments replies endpoint, and resolve threads with the resolveReviewThread mutation.

How to drive a pull request to approved status automatically?▼

Run repeated passes that observe reviewDecision, CI checks, and mergeable state via gh pr view and gh pr checks, react to any new non-self review by verifying and addressing findings, and loop until a fresh APPROVED review lands on the current head commit.

Does this Skill merge the pull request after approval?▼

No, it never merges. The loop ends once the PR reaches a fresh APPROVED review with all CI green and mergeable status, then it notifies the user that a human must perform the merge.

Can it handle false positive review findings from bots?▼

Yes. Every finding is verified against the actual code at the cited file and line before acting. False positives receive an evidence-based reply explaining why the finding does not hold, without changing any code.

What happens when a red status comes from a review bot instead of CI?▼

The Skill distinguishes a bot-posted commit status, which is the bot's approve or reject verdict, from real CI checks like build, test, and lint. Each red signal is diagnosed by its source before any fix is attempted.

Why did a re-run of a CI check fail without code changes?▼

A gh run rerun re-evaluates live gates against current external state, so transient infrastructure drift can flip results unrelated to your diff. The Skill checks what actually failed before assuming the diff caused it and never mutates shared external state to force a gate green.

Related Skills

Explore More Software Engineering Skills→
fxfx

pr-preparer

Prepares and opens ready-for-review GitHub pull requests from completed branch changes.

Community
Advanced
adryanmoldokkr32-pixeladryanmoldokkr32-pixel

land-and-deploy

Merge pull requests, monitor CI, deploy to production, and verify health with canary checks.

Community
Intermediate
maximhqmaximhq
7.7k

review-pr

Reviews pull request diffs with parallel finder agents and adversarial verification to produce findings and merge recommendations.

Official
Advanced
narduk-enterprisesnarduk-enterprises

gh-copilot-pr-loop

Automate GitHub Copilot PR review, fix, and merge cycles.

Official
Advanced