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.
npx skills add DJRHails/dotfiles --skill review-prgit clone --depth 1 https://github.com/DJRHails/dotfilesWrote 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/djrhails/dotfiles/review-pr)<a href="https://agentmods.dev/skills/djrhails/dotfiles/review-pr"><img src="https://agentmods.dev/badge/skills/djrhails/dotfiles/review-pr/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/djrhails/dotfiles/review-pr"><img src="https://agentmods.dev/badge/skills/djrhails/dotfiles/review-pr.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 | $0.00087 | $0.03353 |
| Opus 5.5 | $0.00035 | $0.01341 |
| Sonnet 5.5 | $0.00017 | $0.00671 |
| Haiku 4.5 | $0.00009 | $0.00335 |
Grade A, and why
review-pr 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 — 248 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Review and Fix PR
$ARGUMENTS is the PR number (e.g. /review-pr 42). Without one, ask which PR.
Every finding must be verified before it is posted (a code path you read or ran), and P1/P2 findings are fixed on the PR branch. Those two properties are the product. Everything else in this file is about doing them in as few API requests and as little context as possible: a review's cost is the number of requests times the context each one re-reads, so the rules below are shaped as mechanisms (what to run, what to read, what to keep out of the transcript), not exhortations.
Ground rules
- Canonical repo.
upstreamremote if present, elseorigin; pass--repo <owner/name>on everyghcall. - Shared-repo guard. Solo repo = owner
DJRHailsandgh api repos/<owner>/<repo>/collaborators --jq lengthreturns 1: act autonomously. Any other repo: summarise what you are about to post/push and wait for an explicit yes before each action that reaches other people (§3 review, §4 push/resolve, §5 summary). As a gantry worker the spawn is the authorization — never block. - Gantry sign-off. When
$GANTRY_URLis set, end every body you post (review body, each inline comment, each thread reply, the summary) with_[via gantry](<GANTRY_URL>)_as its last line. It is the loop guard portcullis keys on to drop your own posts when they echo back as webhooks — an unsigned reply re-triggers your own run (touchstone#2290: 11 unsigned replies, 11 phantom re-reviews). - Context discipline (these bound cost without touching what you review):
- One message per batch of independent calls: metadata + diff stat + file list + collaborator count is one message, not five.
- Read the diff once, from disk. Read files by line range around the
hunks (
sed -n 'a,bp'), notcatof whole files; a whole-file read only for files under ~300 lines. - Every command that can print more than ~150 lines writes to a file
(
> /tmp/review/<name>.log 2>&1) and you read atail -40/rgof it — never the whole output into the transcript. - Keep any single tool call under ~4 minutes (
timeout 240, or run in the background and poll at ≤4-minute intervals): the 5-minute prompt cache expires across a longer gap and the next request re-writes the entire context at write rates. - Narration between calls: one line when a decision needs recording.
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 Changed · -167 lines · -2 tokens per session 0f6b994aae8e
- 1mo ago First seen · 415 lines · 89 tokens per session scan A 4aff699b1571
review-pr is a skill published in the GitHub repository DJRHails/dotfiles (2 stars, last pushed yesterday), licensed MIT. It adds 87 tokens to every session and 3,353 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 skills, from other repositories
verify-behavior
Verify or reproduce visible product behavior by driving the real UI with pi-computer-use's checked tools, requiring verified expect postconditions and durable state evidence for meaningful UI flows. Use when triage needs visual reproduction, implementation needs behavioral proof, review needs interactive confirmation…
source-drafting
Use when drafting documents, reports, posts, presentations, or review comments from source materials, or conducting research across provided documents. Enforces read-first drafting, a source map for every claim, agent verification, and separation of verified from unverified claims.
autopilot
Carry a well-scoped GitHub issue through the full dev loop autonomously, stopping at a per-run tier boundary (PR-ready, or merge+deploy for small reversible changes).
autopilot-triage
Vet open issues for autonomous resolution and queue the qualifying ones with the autopilot-queued label — the start-of-day "fill the queue" half of the triage → run split.
create-pr
Create a pull request with auto-generated description, issue linking, ROADMAP updates, and PR-metadata validation.
drift-check
Pre-PR advisory check for deviations from the project spec. Read-only analysis.