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 j4rk0r/claude-skills --skill codex-pr-reviewgit clone --depth 1 https://github.com/j4rk0r/claude-skillsWrote 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/j4rk0r/claude-skills/codex-pr-review)<a href="https://agentmods.dev/skills/j4rk0r/claude-skills/codex-pr-review"><img src="https://agentmods.dev/badge/skills/j4rk0r/claude-skills/codex-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/skills/j4rk0r/claude-skills/codex-pr-review"><img src="https://agentmods.dev/badge/skills/j4rk0r/claude-skills/codex-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 | $0.00113 | $0.03873 |
| Opus 5 | $0.00056 | $0.01937 |
| Sonnet 5 | $0.00023 | $0.00775 |
| Haiku 4.5 | $0.00011 | $0.00387 |
Grade A, and why
codex-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 11d 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 — 223 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Revisión Codex — Pull Request
Primera línea del informe generado: "Español confirmado."
Antes de revisar — pregúntate (framework Codex)
- ¿Qué tipo de cambio es? Hook nuevo, refactor, hotfix, migración, config — el tipo determina qué puntos Codex aplican.
- ¿Cuál es el peor escenario en producción? Si este código falla, ¿qué se rompe? Eso fija la severidad de los hallazgos.
- ¿Hay algo fuera del diff que el cambio asume? Schema, config, dependencias, índices BD, permisos — los olvidos viven en lo que no se ve.
- ¿Es idempotente? Si se ejecuta dos veces (retry, doble clic, re-deploy, re-import config), ¿pasa algo malo?
- ¿Se puede desactivar? ¿Hay kill-switch (config/setting/feature flag) si la feature explota a las 3am sin tiempo de redeploy?
Si no puedes responder a las cinco con confianza, lee el código circundante antes de emitir el informe.
Ejemplo de aplicación (mini-PR)
PR: añade mymodule_node_update() que calcula un score y lo guarda en una tabla custom vía db_query("INSERT ... VALUES ('" . $title . "')").
- (1) Tipo: hook de entidad + escritura SQL → aplican Codex 1, 3, 4, 9.
- (2) Peor escenario: SQL injection si
$titleviene de input + nodos nuevos sin score (falta_insert). - (3) Asume: que la tabla custom existe (¿hay update hook? ¿schema?).
- (4) Idempotente: ¿qué pasa al re-guardar el nodo? ¿duplica filas o hace UPDATE? Verificar.
- (5) Kill-switch: no hay → hallazgo Media (Codex 11).
Resultado: 4 hallazgos (3 Alta + 1 Media), ninguno fuera del alcance del PR.
Fast path (resumen ejecutable)
1. Leer references/metodologia-codex-completa.md (~70 líneas, completo)
2. Leer references/plantillas-hallazgos.md (~230 líneas, completo)
3. cd a drupal/ si existe, si no raíz del workspace
4. Pedir/confirmar número de PR y ramas (base ← head). Si no se sabe, preguntar.
5. git fetch origin pull/<N>/head:pr-<N> (descarga rama del PR)
6. git diff --name-only origin/<base>...pr-<N> → lista de archivos
7. Aplicar Decision tree para elegir puntos Codex prioritarios
8. Aplicar las 5 preguntas del framework Codex (sección anterior)
9. Revisar archivo por archivo, anotando hallazgos con Severidad
10. Detectar IDE: leer CLAUDE_CODE_ENTRYPOINT (claude-vscode/claude-cursor/claude-antigravity).
Solo si no es concluyente, caer a detección por carpeta existente.
11. Escribir informe en <carpeta-IDE>/Revisiones PRs/lint-review-pr<N>.md
12. Auto-verificar contra el Checklist de auto-verificación (final del documento)
What ships with it
9 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 11d ago First seen · 223 lines · 113 tokens per session scan A c57072894f2f
codex-pr-review is a skill published in the GitHub repository j4rk0r/claude-skills (2 stars, last pushed 1mo ago), licensed MIT. It adds 113 tokens to every session and 3,873 once invoked, about $0.0006 per session on Opus 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
la-chancla
Use only when the user explicitly requests a strict review of one named task, change, plan, release, or invariant. Never self-selected, never auto-loaded, never invoked by another skill or by the mexican-mom router. NOT a session-wide mode and NOT a warning counter.
frijoles-en-el-tupper
Use when writing or reviewing names — functions, variables, types, files, schemas, API fields, config keys — and the comments attached to them, to check that each label predicts what is actually inside. NOT for whether failure paths are handled; use pero-ponte-sueter.
pero-tu-primo
Use after a substantial implementation works and before claiming it is good, clean, idiomatic, or optimal. Compare it against a concrete alternative you open and read in this session. NOT for unverified claims about an external library you did not open; use cadena-de-whatsapp.
si-el-lo-hace-tu-tambien
Use before giving your own technical justification when it rests mainly on popularity, trend, convention, or what another codebase does rather than on the requirements at hand. Name the requirement and the trade-off instead. NOT when the user already chose the tool; use porque-yo-lo-digo.
shipkit-review-shipping
Review changes across 12 quality dimensions and report findings. Use after a chunk of work or before commit.
shipkit-prompt-audit
Audit LLM prompt pipeline architecture — decomposition, parallelization, chain integrity, schema validation, fallback paths. Finds structural issues no linter catches.