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/atretyak1985/swarmeryWrote 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/commands/atretyak1985/swarmery/pr-review)<a href="https://agentmods.dev/commands/atretyak1985/swarmery/pr-review"><img src="https://agentmods.dev/badge/commands/atretyak1985/swarmery/pr-review/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/commands/atretyak1985/swarmery/pr-review"><img src="https://agentmods.dev/badge/commands/atretyak1985/swarmery/pr-review.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 | $0.00039 | $0.03179 |
| Opus 5.5 | $0.00016 | $0.01272 |
| Sonnet 5.5 | $0.00008 | $0.00636 |
| Haiku 4.5 | $0.00004 | $0.00318 |
Grade A, and why
pr-review 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 14d 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 — 248 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/pr-review — review a PR, or triage the comments on it
Arguments: $ARGUMENTS — <pr-url | repo#number> [review|triage].
Two modes, one procedure family:
- triage — there are reviewer comments (a bot or a human) on the PR. Verify each one against the real code, fix the valid ones on the PR branch, and reply in the comment thread with a verdict backed by evidence.
- review — no comments to triage (or explicitly requested). Review the PR diff yourself and post verified findings as one review with inline comments.
Step 0 — load the project's review config
Read ${CLAUDE_PROJECT_DIR}/.claude/project.json → prReview:
owner— the GitHub org or user that owns the repos.worktreeRoot— where throwaway worktrees go (default: apr-review/dir under$AGENT_WORKSPACE_ROOT, else$TMPDIR).repos.<name>— per repo:path(the local main checkout),setup(commands to run in a fresh worktree),gates(commands that must pass before anything is pushed),reviewSkills(skills to load before a review),notes(free-text quirks).knowledge— optional path to a project-local markdown file with verification heuristics, known CI flake signatures and anything else a reviewer must know. Read it in full when present; it overrides nothing below, it adds to it.
Stop with a clear message if prReview is missing or the PR's repo is not under
repos — never guess a gate. Resolve <repo> from the PR URL or repo#number.
Mode auto-detection when not given: if
gh api repos/<owner>/<repo>/pulls/<n>/comments --paginate returns unresolved
top-level comments (in_reply_to_id == null) authored by someone other than the PR
author that have no reply yet → triage; otherwise → review.
Non-negotiable rules
- Verify before believing — in both directions. Never implement a reviewer
suggestion and never post a finding without checking it against the code first.
No "You're absolutely right", no thanks, no praise. A wrong rebuttal is worse than
a fixed non-bug, so the evidence bar for "INVALID" is a concrete guard, constraint,
or test — cited as
file:line. - Never touch the main checkouts.
repos.<name>.pathmay hold the user's WIP on another branch. All work happens in a throwaway worktree. - Thread replies, not top-level comments. Replies go to
gh api repos/<owner>/<repo>/pulls/<n>/comments/<id>/replies -f body=.... A top-levelgh pr commentbreaks the thread and a review bot re-flags the finding. - Idempotency. Before replying, list the thread's existing replies; if the PR author (or this tool) already answered a comment, skip it.
- Scope discipline. A real problem that predates the PR (visible via
git diff origin/<base>...HEAD) is mentioned in the reply ("predates this PR, left for a separate cleanup") — not fixed in the PR. - Nothing is pushed and no reply is posted until the repo's
gatesare green. A gate that cannot run (no test runner configured, say) is reported as exactly that — never as "tests pass".
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.
- 14d ago First seen · 248 lines · 39 tokens per session scan A 81d44822d3fc
pr-review is a command published in the GitHub repository atretyak1985/swarmery (7 stars, last pushed 2d ago), licensed Apache-2.0. It adds 39 tokens to every session and 3,179 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 2026-10-01.
Other commands, from other repositories
devkit.github.review-pr
Provides comprehensive GitHub pull request review with code quality, security, and best practices analysis. Use when reviewing a PR before merging.
hatch3r-pr-resolve
Read open PR comments, evaluate each against current code via the rigor contract, implement accepted findings, reply inline. Multi-platform.
create-pr
Context-aware GitHub Pull Request creation and update assistant.
commit-and-create-mr
Commit and create MR on Gitlab, with post-merge worktree cleanup.
pr
Analyze changes, detect scope issues, and create a well-structured PR.
review
Recent git commits: !git log --oneline -10.