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 agentmods add instructions/figitaki/reviews/agents-mdgit clone --depth 1 https://github.com/figitaki/reviewsWrote 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/instructions/figitaki/reviews/agents-md)<a href="https://agentmods.dev/instructions/figitaki/reviews/agents-md"><img src="https://agentmods.dev/badge/instructions/figitaki/reviews/agents-md.svg" alt="Measured on agentmods" 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 | $0.07046 | $0.07046 |
| Opus 5 | $0.03523 | $0.03523 |
| Sonnet 5 | $0.01409 | $0.01409 |
| Haiku 4.5 | $0.00705 | $0.00705 |
Grade C, and why
reviews AGENTS.md scanned grade C with 1 finding 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 4d 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
The repo's `.claude/settings.local.json` allows broad `mix`/`cargo`/`npm`/`bun`/`git add`/`git commit`/`reviews push` but denies destructive git operations (`push` is allowed to `origin` but the deny list blocks `reset`, How it starts
The opening of the file, as written. The whole thing — 504 lines — stays where its author put it; the contents beside it link to each section on GitHub.
This is Reviews — a code-review tool for arbitrary diffs (not just GitHub PRs). Stack: Phoenix 1.8 + LiveView (web), Rust CLI (cli/), Postgres, @pierre/diffs React island for diff rendering.
Working in this repo
Dev server
Use ./bin/server — it sources .env.local (your GITHUB_CLIENT_ID / GITHUB_CLIENT_SECRET) then execs mix phx.server on :4000. Don't run mix phx.server directly unless you're certain you don't need OAuth.
Don't start a second Phoenix server while one is already running on :4000 — bind clash. Check first: lsof -iTCP:4000 -sTCP:LISTEN.
Tests
mix test— full suite (currently 27 tests, all should pass).mix compile --warnings-as-errorsbefore committing.mix format --check-formatted.- Postgres test DB is at
reviews_test;pg_isreadyto check before running.
Pushing diffs to view as reviews
The Rust CLI at cli/target/release/reviews is how you preview your own changes: cd <any-git-checkout> && reviews push posts the current branch's diff to the local Phoenix server and prints a http://localhost:4000/r/<slug> URL. reviews push --update <slug> adds a patchset to an existing review.
Git remote
Whichever git remote the maintainer has configured as origin is the canonical one. Don't add other remotes without checking; don't git push anywhere but origin unless explicitly asked. If a github remote is also configured, it's a mirror that's pushed to as part of the release process — see docs/RELEASE.md.
Plans and memory
.plans/— project-scoped design plans, tracked in git. Start here when picking up work: read any plan that references the area you're touching..plans/a11y-design-fixes.mdis the current open plan (P0/P1/P2 a11y items, two flagged⚠ overlapwith the deferred Shiki swap)..claude/— per-checkout Claude Code state (settings, worktrees, transcripts). Gitignored. Don't put anything here you want shared.- Auto-memory at
~/.claude/projects/-Users-warbler-src-reviews/memory/captures user-level feedback (e.g. "brainstorm UX before plans," "don't self-grant permissions"). ReadMEMORY.mdfirst if you have access.
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.
- 4d ago First seen · 504 lines · 7,046 tokens per session scan C 695187726332
reviews AGENTS.md is an instructions file published in the GitHub repository figitaki/reviews (6 stars, last pushed 11d ago), licensed MIT. It adds 7,046 tokens to every session, about $0.0352 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
jido_managed_agents AGENTS.md
AGENTS.md instructions for agentjido/jido_managed_agents, covering project guidelines, phoenix v1.8 guidelines, js and css guidelines, ui/ux & design guidelines and release hygiene.
jido AGENTS.md
AGENTS.md instructions for agentjido/jido, covering agents.md - jido guide, intent, runtime baseline, commands and architecture snapshot.
jido_ai AGENTS.md
Instructions for agentjido/jido_ai, covering agents.md - jido.ai guide, intent, runtime baseline, commands and architecture snapshot.
lang_ex AGENTS.md
Instructions for freshaengineering/lang_ex, covering langex, commands, architecture, behaviours and key design decisions.
excessibility CLAUDE.md
Claude Code instructions for lessthanseventy/excessibility, covering claude.md, project overview, development commands, testing and run all tests.
opal AGENTS.md
Instructions for matteing/opal, covering opal project guide, architecture, components, built-in tools and development.