appsec
Skill0 starsChanged 24 days agoMITView on GitHub
Code reviewer with one entry point: /appsec. Use whenever the user wants code reviewed: current-branch work including uncommitted and new files ('review my changes', 'review this branch'), a Merge/Pull Request by number or URL ('review MR 123'), several MRs, or a whole repo/directory ('audit the whole repo', 'pre-pentest sweep'). Auto-detects scope and risk, no questions; runs two equal tracks: SECURITY (OWASP/CWE: injection, access control, auth/session, SSRF, XSS, deserialization, crypto, DoS, race, secrets, supply chain, CI/CD, mobile MASVS, LLM/agent/MCP, domain hunting (HTTP smuggling, OAuth/SAML, tenant isolation, cloud IAM, native memory); invariant-based hunting, 5 verification gates, 'needs validation' verdict) and CODE (correctness, business logic vs spec, migrations/API compatibility, performance, over-engineering, tests, conventions; 4 gates). Large or risky changes get the deep engine: recon, coverage table, parallel hunters, independent verifiers. Returns a merge verdict.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- TelosTechnology/skillstheir own version, first seen Jul 17, 2026
- joshuaporth/appsec-skilltheir own version, first seen May 15, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- TelosTechnology/skills★ 1same name
- joshuaporth/appsec-skill★ 0same 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.
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 20
If the user asks a single question about secure coding, severity, or CWE instead of requesting a review: answer briefly from the …
- Distinctive: line 24
Read only the file you need, not all of them.
- Distinctive: line 48
Commit the lockfile; dependencies and base images come from controlled sources.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
appsec
One purpose: code review. One command: /appsec. Everything after it is optional plain language, not flags:
| Goal | Example |
| Work in progress on this branch (including uncommitted, new files) against the target branch | /appsec |
| MR/PR by number or link | /appsec 123, /appsec !123, /appsec <URL> |
| Anything else | /appsec against develop, /appsec uncommitted only, /appsec be thorough, /appsec audit the whole repo, /appsec src/billing, /appsec 123. Spec: ... |
The skill decides scope, standard or deep review, which domains to hunt, and the independent reviewer on its own; it does not ask again unless it cannot continue. It always reviews two equal tracks: Security and Code (logic, business rules, design, over-engineering, performance, tests, conventions). Respond in the user's language (default English); keep code and identifiers in English.
Follow workflows/review.md. Deep review (large/risky diff, whole repo, or the user says "thorough") automatically moves the Security track to workflows/review-deep.md.
If the user asks a single question about secure coding, severity, or CWE instead of requesting a review: answer briefly from the references, citing the specific section.
References
Read only the file you need, not all of them.
| File | Content |
| references/code-review.md | Code track: BLOCKER/MAJOR/MINOR/NIT severity, 4 gates, intent sources, 9 hunting groups, over-engineering table |
| references/secure-coding-patterns.md | Security: wrong, bypass, correct fix; known-safe |
| references/hunting-method.md | Invariant-based hunting, six trust-boundary-outcome points, Needs validation verdict, hardening, safe verification, domain selection table, feature abuse, vulnerability chaining, wildcard, anti-patterns |
| references/domains/ | Domain hunting: web-auth, client-side, data-isolation, resource-exhaustion, protocols-messaging, cloud-deployment, ci-release, local-ipc-desktop, native-memory. Open only the file whose boundary the diff/repo touches |
| references/severity-cwe.md | Severity anchor scale, severity table, CWE; "not a security finding" |
| references/change-classification.md | 10 change groups, sensitive code areas, signal-collecting commands |
| references/supply-chain-devsecops.md | Lockfile, dependency sources, CVE, SBOM, CI tokens, pipeline |
| references/secure-design.md | Cross-checking a large MR or new integration against security design notes, authz matrix, acceptance criteria |
| references/ai-llm-security.md | OWASP LLM Top 10 2025, lethal trifecta, hunting RAG/memory/agent/MCP bugs |
| references/mobile/ | lessons.md, android.md, ios.md, cross-platform.md |
| references/report-template.md | Report template |
| references/multi-mr.md | MR by IID (GitLab/GitHub), multiple MRs, multiple repos |
| references/tob.md | Trail of Bits plugin (optional) |
Fixed principles
- Every finding must have
file:lineevidence and pass the verification gates of its track (Security 5 gates, Code 4 gates). A Security finding names the less-trusted principal, the control that should have blocked it, the boundary crossed, and the concrete outcome. A serious finding must pass adversarial challenge, ideally by a different subagent from the one that found it. Never report on feeling or taste. - Few and correct beats many and noisy. Unsure about business rules: ask a question, do not conclude. Depends on facts outside the repo (proxy, IdP, IAM, ingress, browser): record Needs validation, assign no severity, do not guess either way. Missing a defense layer when another layer already blocks it: hardening, not a finding.
- Controls are decided on the server; identity, tenant, role come from a verified token. Allowlist over blocklist. Fix the root cause.
- Simple is the default: every abstraction must serve a current requirement; a design proposal always comes with a simpler version.
- Commit the lockfile; dependencies and base images come from controlled sources. Secrets stay out of code, images, logs, prompts; once on the remote it is exposed: rotate.
- Block the merge only for issues caused by this change; pre-existing bugs go in the Context section.
- The report talks only about code: no who ran it, machine, time, email, internal URL.
- The review is read-only and leaves no trace in the project: do not edit code, create files, config, hooks, refs, worktrees (only
git fetchupdating remote refs is fine), do not checkout, stash or change branches. Do not fix code or post comments unless the user asks. Files that must be saved (report, coverage) go in~/.claude/appsec-reviews/. - Verification is source-based; do not probe deployed environments, external services, or real data or credentials. Run repo code only when needed to settle a candidate, with fake data, no network, no dependency installation; if that cannot be guaranteed, do not run it and record Needs validation.
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.