Borrowing it
Nothing to install: this file belongs to mindroom-ai/mindroom. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/mindroom-ai/mindroom/main/.claude/skills/native-subagent-pr-loop/SKILL.mdgit clone --depth 1 https://github.com/mindroom-ai/mindroomWrote 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/mindroom-ai/mindroom/native-subagent-pr-loop)<a href="https://agentmods.dev/skills/mindroom-ai/mindroom/native-subagent-pr-loop"><img src="https://agentmods.dev/badge/skills/mindroom-ai/mindroom/native-subagent-pr-loop/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/skills/mindroom-ai/mindroom/native-subagent-pr-loop"><img src="https://agentmods.dev/badge/skills/mindroom-ai/mindroom/native-subagent-pr-loop.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00042 | $0.01088 |
| Opus 5.5 | $0.00017 | $0.00435 |
| Sonnet 5.5 | $0.00008 | $0.00218 |
| Haiku 4.5 | $0.00004 | $0.00109 |
Grade A, and why
native-subagent-pr-loop 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 19d 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 — 110 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Native Subagent PR Loop
Core Rule
The main thread owns all repository mutations. Native subagents provide bounded analysis or review, but review findings are untrusted until the main thread verifies them against current code. After each fix, use fresh read-only reviewers with neutral prompts.
Use When
- The user asks for native Codex agents, native sub-agents, parallel review agents, or a review loop.
- The user wants fixes made inline in the main thread after agent findings.
- The work is on a PR or branch where commits and pushes should remain inspectable.
Do Not Use
- The user explicitly asks for
agent-cli, tmux supervision, or another external agent runner. - The task has no useful independent subtask or review surface.
- A reviewer must use private context, CI state, or browser state unavailable to a fresh subagent.
Main Loop
- Pin context.
Read repo instructions, check
git status --short --branch, identify base/head, and keep the active branch unless a separate branch is clearly safer. In this repo, never createcodex/...branches. - Implement in the main thread.
The main thread edits, verifies, stages targeted files only, commits, and pushes.
Do not use
git add ., amend, or force-push unless the user explicitly asks. - Use task subagents only for bounded, non-overlapping analysis or review. Subagents must not edit repository files. The main thread applies any verified changes.
- After each pushed step, start two fresh native Codex PR-review agents.
Use read-only prompts, attach
pr-reviewwhen reviewing merge readiness, and tell them not to edit, commit, push, or inspect CI if the user excluded CI. - Treat findings as claims.
Verify each finding against current code before editing.
Fix only real, in-scope issues in the main thread.
Classify stale, incorrect, overreaching, or duplicate findings instead of patching blindly.
Decline edge cases, as
pr-reviewdefines them, with a one-line reason. Record out-of-scope findings aspr-reviewdescribes and do not fix them in this PR. - Repeat after any fix. Commit and push the main-thread fix, close old reviewers, then launch fresh reviewers against the new head. After every third review round that still finds many issues or a new major bug class, stop patching and reconsider the design before another patch round.
- Stop only when both fresh reviewers approve the same head, or when every remaining finding on that head is declined with a recorded reason. Confirm the worktree is clean and the remote branch matches the local head.
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.
- 19d ago Changed c0c8c8886f0d
- 20d ago Changed · +4 lines 7648359fbead
- 1mo ago First seen · 106 lines · 42 tokens per session scan A b41aa9472f16
native-subagent-pr-loop is a skill published in the GitHub repository mindroom-ai/mindroom (346 stars, last pushed today), licensed Apache-2.0. It adds 42 tokens to every session and 1,088 once invoked, about $0.0002 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 2026-08-30.
Other skills, from other repositories
github
GitHub via gh CLI: PRs, issues, reviews, repos, auth.
otto-pr-review-loop
Run Polly and OCR on a PR until all findings settle.
resolve-drive-pr
Drive an open PR through Polly/OCR, validation, and maintainer handoff.
github
GitHub operations via gh CLI: issues, PRs, CI runs, code review, API queries. Use when: (1) checking PR status or CI, (2) creating/commenting on issues, (3) listing/filtering PRs or issues, (4) viewing run logs. NOT for: complex web UI interactions requiring manual browser flows (use browser tooling when available)…
issue-to-pr-resolver
Implement a GitHub issue end-to-end — create a worktree branch, implement the feature with tests, create a draft PR, then iteratively resolve all CI failures and review comments until the PR is clean. Use when you need to fully implement a GitHub issue from start to merge-ready. Triggers on "implement issue", "resolve…
my-pr-checker
Manage your own GitHub pull requests — check CI status, inline review comments, PR-level comments, resolve review threads, fix issues, and iterate until all checks pass and threads are resolved. Use for managing your own PRs (not external contributions). Triggers on "check my PR", "check PR", "/my-pr-checker ".