sec-review
Skill1 starsChanged 35 days agoMITView on GitHub
Diff-scoped security audit that runs in an isolated fork and returns its findings as a result, so a chaining caller (/super-review) keeps control. Proves a non-empty diff first, then audits every changed file (committed, uncommitted, untracked) for vulnerabilities the change introduces. Standalone it is chain-safe too. NOT /code-review (correctness) or /blast-radius (downstream consumers).
3 other projects use a version of this.
Adopted by
Most recently taken up by
- Asno-dev/apextheir own version, first seen Oct 15, 2026
- NeptuneHub/AudioMuse-AItheir own version, first seen Oct 8, 2026
- parameshjava/anthropic-claude-codetheir own version, first seen Sep 20, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- NeptuneHub/AudioMuse-AI★ 3ksame name
- Asno-dev/apex★ 1same name
- parameshjava/anthropic-claude-code★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository56 other files it gives its agents.
- autopilotSkill
- before-afterSkill
- bigpictureSkill
- blast-radiusSkill
- brainstormSkill
- compoundSkill
- confidence-checkSkill
- contextualizeSkill
- create-ship-skillSkill
- create-skillSkill
- curateSkill
- debugSkill
- deep-researchSkill
- devtoolsSkill
- distillSkill
- done-checkSkill
- doneSkill
- double-checkSkill
- envsetSkill
- explainSkill
- fortifySkill
- goSkill
- handoffSkill
- harness-auditSkill
- interviewSkill
- matrixSkill
- mockupSkill
- mutateSkill
- notesSkill
- postmortemSkill
- presentSkill
- qaSkill
- readableSkill
- recommendSkill
- reportSkill
- retireSkill
- review-planSkill
- scoutSkill
- setup-ticketingSkill
- setup-worktreesSkill
- spec-checkSkill
- specSkill
- stepsSkill
- successionSkill
- super-reviewSkill
- tangentSkill
- tddSkill
- tend-memorySkill
- testSkill
- ticketSkill
- tldrSkill
- triageSkill
- waitSkill
- workSkill
- worth-itSkill
- wrap-upSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 49
Test files and regex DoS are in scope — don't skip them as "low signal".
- Distinctive: line 58
Do not abridge or summarize away detail.
- Distinctive: line 62
A non-empty diff is a precondition, not a nicety: Never return a clean verdict you cannot tie to specific changed lines.
- Very specific: line 64
Don't touch data files: Never open or quote the contents of customer-data files (*.csv, *.xlsx, dumps, *.sql data scripts, exported …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Audit the current change for security vulnerabilities in this isolated fork, then return the findings to the caller.
This skill does not call the native /security-review. That pass computes its diff once, from the session's cwd against origin/HEAD, committed changes only — so when cwd is a parent directory rather than the repo holding the change, on uncommitted work, or when the branch was cut from something other than the default branch, it reviews the wrong diff or none. The audit below is the review.
What to do
0. Establish the diff — and PROVE it is non-empty (do this first, always)
Find the repo. The fork's cwd may be a parent directory that holds none of the change. Resolve the target repo from $ARGUMENTS or the caller's named worktree/path, falling back to cwd; then run every git command as git -C <repo>.
git -C <repo> rev-parse --show-toplevel # the repo root, not a parent dir
git -C <repo> rev-parse --abbrev-ref HEAD
git -C <repo> diff --stat <base>...HEAD ; git -C <repo> diff --stat # committed + uncommitted
git -C <repo> status --porcelain # UNTRACKED too — git diff never shows new files<base> is the branch this work was cut from (the project's integration branch), not origin/HEAD — that is only the remote's default branch, which may not be where this work started, and diffing against it drags in commits this change never touched.
The untracked line is not optional. Neither git diff form sees a file that has never been added, so an all-new-files change reads as an empty diff. Count a file as part of the diff if it appears in any of the three commands.
Record the file list and total changed lines. Two hard rules:
- If the diff is empty, STOP and return
🛑 could not establish a diff— naming the cwd, the resolved toplevel, and the base you compared against. Do NOT return "clean". An empty diff is a tooling failure, not a security result. - If
$ARGUMENTSnamed a scope, confirm the diff actually contains it. A scope hint that matches nothing is the same failure in a quieter costume.
1. Audit the diff
Review exactly the step-0 file list: the diff hunks for tracked files, and every untracked file read in full. Read surrounding code where it decides the answer — e.g. how sibling resolvers/routes gate the same data. If $ARGUMENTS names a focus, go deepest there, but still cover every file.
Audit as a senior security engineer. Check at least:
- Authorization — every new or changed resolver, route, job, or query gated like its siblings; object-level checks (can this caller see this row?), not just "is logged in".
- Authentication and tokens — signature,
aud,iss, and expiry actually validated; sessions and cookies scoped correctly. - Injection and construction — SQL, shell, template, path traversal, SSRF, unsafe deserialization.
- Input handling — regexes that backtrack on crafted input (ReDoS), unbounded sizes or loops on caller-controlled data.
- Secrets and data exposure — secrets or crypto misuse in code/config; personal data in logs, error messages, API responses, or test fixtures.
- Dependencies — new or bumped packages in the auth/parse/network path.
Test files and regex DoS are in scope — don't skip them as "low signal". Flag only what this diff introduces or newly exposes; a pre-existing issue is worth a line only when the change makes it reachable.
2. Return the findings
Return the complete findings verbatim — each with file:line, severity, and a concrete attack path (who can do what, how). Then:
- Coverage:
reviewed N/N step-0 files— name any file you did not review and why. A partial review is never "clean". - Verdict: clean / N findings. If there are no findings, say so plainly.
Do not abridge or summarize away detail; the caller needs the actual findings to triage.
Guardrails
- A non-empty diff is a precondition, not a nicety. Never return a clean verdict you cannot tie to specific changed lines.
- Read-only. Surface findings; do not edit code. Fixing is the caller's decision.
- Don't touch data files. Never open or quote the contents of customer-data files (
*.csv,*.xlsx, dumps,*.sqldata scripts, exported reports) — review code and config, not data.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

