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 skills/jaansokk/cursor_tools/code-review-codexnpx skills add jaansokk/cursor_tools --skill code-review-codexgit clone --depth 1 https://github.com/jaansokk/cursor_toolsWhat 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 | $0.00045 | $0.01622 |
| Opus 5 | $0.00023 | $0.00811 |
| Sonnet 5 | $0.00009 | $0.00324 |
| Haiku 4.5 | $0.00005 | $0.00162 |
Grade A, and why
code-review-codex 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.
How it starts
The opening of the file, as written. The whole thing — 157 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Code Review via Codex CLI
A read-only, post-implementation review using OpenAI Codex CLI as an independent second opinion. The goal is to catch real issues before commit, not to make edits.
When to use
- After completing a feature, fix, or refactor — as the last step before committing
- When the user says "review", "check my code", "final review", or "sanity check"
- For security or correctness confidence on sensitive code paths
Behavior
- Read-only: Codex reviews and reports. No files are modified.
- Report back: Output is returned to Claude as structured findings for the user to triage.
- Skip noise: No style nits, no formatting opinions, no "consider renaming" suggestions.
Steps
-
Determine scope from
$ARGUMENTS:- No arguments → review uncommitted changes (
--uncommitted) - Branch name (e.g.
main) → review diff against that branch (--base <branch>)
- No arguments → review uncommitted changes (
-
Run the Codex review using
codex exec(supports full prompt alongside diff scope):
Uncommitted changes (default):
codex exec \
-m gpt-5.3-codex \
-c model_reasoning_effort='"xhigh"' \
-s read-only \
--ephemeral \
-o /tmp/code-review-output.md \
"You are a senior engineer doing a final gate review on code that is about to be committed. This is post-implementation — the work is done, you are looking for real problems only.
Start by running: git diff HEAD to see all uncommitted changes (staged and unstaged). Also run git status to catch untracked files that may be relevant.
Review ONLY for issues that would cause bugs, outages, security incidents, or serious perf regressions in production. Ignore style, naming, formatting, and minor code smells.
Use the context7 MCP to check documentation of libraries when needed.
## Review checklist
### 1. Correctness
- Logic errors: wrong conditions, off-by-one, inverted checks, unreachable branches
- State bugs: stale closures, race conditions, mutations of shared state
- Null/undefined access, unhandled promise rejections, uncaught exceptions
- Incorrect API usage or wrong assumptions about library behavior
- Edge cases: empty inputs, boundary values, concurrent access
### 2. Security
- Injection vectors: SQL, XSS, command injection, path traversal
- Auth/authz gaps: missing permission checks, broken access control
- Secrets or credentials in code, logs, or error messages
- Insecure defaults: permissive CORS, disabled CSRF, weak crypto
- User input flowing unsanitized to sensitive sinks
### 3. Data integrity
- Missing or incorrect validation at system boundaries (API inputs, DB writes, file I/O)
- Silent data loss: swallowed errors, ignored return values, dropped events
- Inconsistent state from partial failures (no transaction, no rollback)
### 4. Performance (only if clearly problematic)
- N+1 queries, unbounded loops over large datasets
- Blocking calls in async/event-loop code
- Missing pagination or limits on user-controlled queries
- Obvious memory leaks (growing caches, unclosed resources, dangling listeners)
### 5. Robustness
- Error handling: catch blocks that swallow context, missing retry/backoff on network calls
- Missing timeouts on external calls (HTTP, DB, queues)
- Failure modes that cascade (one failing dependency takes down everything)
## Output format
For each finding:
- **Severity**: CRITICAL / HIGH / MEDIUM / LOW
- **File:Line**: exact location
- **What**: one-sentence description of the problem
- **Why it matters**: what breaks or goes wrong in production
- **Fix**: concrete code suggestion or approach (keep it short)
Group findings by severity (CRITICAL first). If nothing significant is found, say so — a clean review is a valid outcome.
End with a one-line verdict: PASS (ship it), PASS WITH NOTES (minor items), or BLOCK (must fix before commit)."
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.
- 2d ago First seen · 157 lines · 45 tokens per session scan A b50031235e8d
code-review-codex is a skill published in the GitHub repository jaansokk/cursor_tools (1 stars, last pushed 5mo ago), licensed MIT. It adds 45 tokens to every session and 1,622 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 2026-08-31.
Other skills, from other repositories
continuous-discovery
Guide teams through building sustainable customer interview habits and discovery practices. Use when setting up weekly customer interviews, preparing interview guides, coaching story-based interviewing technique, synthesizing user research findings, planning assumption tests, or helping teams that say they don't have…
prd-writer
Guide users through writing Product Requirements Documents (PRDs) and decomposing them into executable technical tasks. Use when creating a PRD, product spec, product one-pager, feature brief, PRP, or when breaking requirements into tasks with estimates, sprint planning, or technical decomposition.
product-led-growth-playbook
Evaluate growth strategy, growth team structure, and go-to-market motions using Elena Verna's PLG frameworks. Use when the user asks about product-led growth, PLG, growth team hiring, self-serve vs sales-led motions, product-led sales, PQA/PQL models, growth loops, when to hire a head of growth, earned vs rented…
ab-testing-framework
Design, run, and analyze A/B tests (controlled experiments) using Ronny Kohavi's methodology and Gibson Biddle's DHM trade-off analysis. Use when the user needs to plan an experiment, choose metrics (OEC), evaluate statistical significance, assess sample size requirements, avoid common experimentation pitfalls, or…
dhm-strategy-framework
Evaluate and strengthen product strategy using Gibson Biddle's DHM framework (Delight, Hard-to-copy, Margin-enhancing). Use when the user asks about product strategy, competitive advantage, feature prioritization trade-offs, or wants to stress-test whether a product idea is strategically sound.
ai-evals-builder
Build AI evals using the Husain-Shankar framework (error analysis, open/axial coding, LLM-as-judge). Use when a user needs to create, improve, or debug evals for an AI product — including defining failure modes, building LLM judges, or setting up production monitoring for an LLM application.