code-review

A senior code-review workflow for an agent platform, focused on architecture and system boundaries before code details. It produces a structured review of correctness, security, tests, simplicity, and compatibility.

In plain words
What is it for?
Use it when reviewing a pull request, diff, or another agent's change, or when checking your own work before calling it complete.
Why use it?
It catches changes that may work locally but put logic in the wrong layer, weaken security, miss important tests, or break an existing contract.

Skill for Claude CodeCodex

Install

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.

agentmods
npx agentmods add skills/extra-org/extra/code-review
Any agent
npx skills add extra-org/extra --skill code-review
Clone the repo
git clone --depth 1 https://github.com/extra-org/extra

Made for: Claude Code, Codex.

Per session 33 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,144 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00033 $0.01144
Opus 5 $0.00016 $0.00572
Sonnet 5 $0.00007 $0.00229
Haiku 4.5 $0.00003 $0.00114

Measured 2d ago against content hash c17a459c8319, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

code-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 2d 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.

.claude/skills/code-review/SKILL.md · 110 lines

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.

Skill: Code Review

Purpose

Perform senior-level code review on changes in this repository: judge architecture and design first, then correctness, security, testability, and simplicity. The goal is to protect the project's boundaries and quality, not to nitpick syntax.

When to Use This Skill

  • Reviewing a pull request, diff, or another agent's proposed change.
  • Self-reviewing your own change before declaring a task complete.
  • Evaluating whether a change is safe to merge or must be sent back.

Files to Read First

  • AGENTS.md (especially §3 non-negotiable architecture rules).
  • docs/ARCHITECTURE.md and the relevant layer docs (RUNTIME_LIFECYCLE.md, YAML_SPEC.md, PROMPT_RENDERING.md, SIDECAR_CONTEXT_AUTH.md, MCP_AND_TOOLS.md).
  • docs/adr/ for any contract the change touches.
  • .ai/skills/architecture-review.md if the change affects architecture.
  • The task file in tasks/ the change claims to implement.

Core Principles

  • Review architecture before syntax. A clean diff in the wrong layer is still wrong.
  • Boundaries are non-negotiable. Validation, compilation, runtime, prompt rendering, plugin context/access, and tools stay separated.
  • Security is enforced outside prompts. Prompt wording is never a boundary.
  • Simplicity wins. Reject unnecessary abstraction; prefer the smallest design that works.
  • Behavior must be tested, and tests must assert behavior, not internals.

Process

  1. Understand intent. Read the task/PR description and confirm the change is in scope for that task. Out-of-scope churn is a finding.
  2. Architecture & separation of concerns. Confirm code lives in the correct src/agentplatform/<layer> package and does not cross boundaries (e.g. the runtime reading raw YAML, or client/business logic leaking into the runtime).
  3. Lifecycle/state. Verify runtime/startup state is separated from per-request state: nothing request-scoped on RuntimeEngine or the compiled graph; per-request data lives on ExecutionContext.
  4. Public interface. Check that public functions/classes have clear, typed, minimal signatures and stable contracts; private details are not leaked.
  5. Errors. Confirm errors are typed/actionable and name what went wrong (e.g. which YAML key, which missing prompt variable), not bare exceptions.
  6. Security. Confirm permission/tool-policy enforcement happens at the tool/data layer; injected params can't be overridden; no secrets hardcoded; secrets redacted in traces/logs.
  7. Testability & tests. Confirm the code is testable (DI, no hidden globals) and that tests cover behavior and edge/negative cases, not private implementation details.
  8. Simplicity & abstractions. Flag premature abstractions, clever code, and layers collapsed "for convenience".
  9. Backward compatibility. If a public contract (YAML schema, plugin contract, API shape) changed, require an ADR and check for breakage.
  10. Write the structured report (below).

Read the full file on GitHub · 110 lines

Changes

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.

  1. 2d ago First seen · 110 lines · 33 tokens per session scan A c17a459c8319

Subscribe to this mod's changes

code-review is a skill published in the GitHub repository extra-org/extra (108 stars, last pushed 4d ago), licensed MIT. It adds 33 tokens to every session and 1,144 once invoked, about $0.0002 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 2026-08-30.