site-fortress
Skill0 starsChanged 32 days agoMITView on GitHub
Full security audit and hardening of the user's OWN website, web app or browser game, part by part, like a professional pentester — finds every vulnerability and bug, fixes it, and proves the fix. Use before launching any site, when the site handles logins, personal data or payments (cards, Pix), when the user asks "deixa seguro", "verifica a segurança", "procura erros", "testa como hacker", or after big changes. Defensive only — never against sites the user doesn't own.
- Reads credentials
1 other project uses a version of this.
Adopted by
Most recently taken up by
- matheus-rib1108/meu-kit-claude-codeword for word, first seen Oct 17, 2026
Other versions1 project uses a version of this.
Word for word · 1
- matheus-rib1108/meu-kit-claude-code★ 0100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
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 repository28 other files it gives its agents.
- abordagem-whatsapp-vendasSkill
- agent-builderSkill
- assistente-voz-jarvisSkill
- cardapio-digitalSkill
- claude-chatsSkill
- comparador-produtos-brSkill
- cost-estimatorSkill
- curriculo-jovem-aprendizSkill
- github-helperSkill
- humanizar-texto-ptSkill
- legal-complianceSkill
- legenda-instagram-negocioSkill
- matematica-passo-a-passoSkill
- minecraft-construcoesSkill
- orcamento-servicosSkill
- organizador-de-estudosSkill
- pc-controlSkill
- photo-editorSkill
- projeto-senai-desafioSkill
- prospeccao-local-brSkill
- publicar-skills-githubSkill
- revisao-flashcardsSkill
- session-radarSkill
- skill-hunterSkill
- trabalho-escolarSkill
- tutor-sqlSkill
- video-linkSkill
- video-makerSkill
The file
We split this skill into the rules it gives an agent
11 rules. About 0 are common, 6 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 11
Only test systems the user owns or has written permission to test.
- Very specific: line 16
Use parallel subagents for independent phases (e.g.
- Distinctive: line 33
Read routes and config: list every page, API route, server action, webhook, form, file upload, admin area, cron job, third-party script.
- Distinctive: line 35
Write who may call each one (anonymous / logged-in / owner / admin).
- Very specific: line 39
No key in front-end code: in Next.js, anything NEXT_PUBLIC_* is public.
- Very specific: line 48
Use a proven provider (Neon Auth, Auth.js, Clerk, Supabase Auth).
- Distinctive: line 54
Every API route / server action checks on the server who the user is and whether they own the record.
- Distinctive: line 56
Test IDOR: user A changes an id in the URL/body to user B's record → must get 403/404.
- Distinctive: line 68
The site never sees or stores card numbers: Use hosted checkout: Stripe Checkout / Payment Links, Mercado Pago Checkout Pro, or the …
- Very specific: line 100
Check headers: curl -sI <url>.
- Very specific: line 105
Run the legal-compliance skill: privacy policy, terms, cookie consent that really blocks scripts until accepted, consumer info, …
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
Site Fortress
Goal: leave no known hole and no known bug. Be honest with the user: nothing is "100% secure" — the goal is that every known attack class is checked, closed and proven.
Only test systems the user owns or has written permission to test. Active tests (phase 9) run against a preview/staging deploy or localhost, never against someone else's site.
Work through every phase in order. Keep SECURITY-AUDIT.md at the repo root: one row per
finding — ID | fase | severidade (crítica/alta/média/baixa) | arquivo:linha | o que é | como corrigiu | prova.
Use parallel subagents for independent phases (e.g. security-auditor on phases 3–6,
code-reviewer on 7, qa-expert + webapp-testing on 10).
Tool map (all installed in this repo — use them, in parallel where independent)
| Phase | Skills / plugins |
| 0 map | ghost-repo-context, audit-context-building |
| 1 secrets | secrets-gitleaks |
| 2 deps | sca-trivy, supply-chain-risk-auditor |
| 3–7 code | ghost-scan-code (depth=full), cybersecurity, owasp-security, semgrep, sharp-edges, plugin claude-security, agent security-auditor |
| 6 payments | plugins mercadopago / stripe (their review skills) |
| 9 attack | dast-zap, dast-nuclei, webapp-testing, plugin chrome-devtools-mcp |
| 10 legal | legal-compliance (privacy policy, terms, cookies, LGPD) |
| 11 verify | ghost-validate, fp-check, variant-analysis, ghost-report |
Plugin security-guidance is always on: it warns while code is being written. |
Phase 0 — Map the attack surface
- Read routes and config: list every page, API route, server action, webhook, form, file upload, admin area, cron job, third-party script.
- Write who may call each one (anonymous / logged-in / owner / admin). This table drives phases 3–4.
Phase 1 — Secrets
gitleaks detect --source . --log-opts="--all"(whole git history) ortrufflehog git file://..- No key in front-end code: in Next.js, anything
NEXT_PUBLIC_*is public. Server-only keys stay in Vercel/hosting env vars..env*in.gitignore. - Any leaked secret: rotate it at the provider (deleting the commit is not enough).
Phase 2 — Dependencies and supply chain
npm audit --omit=dev/pnpm audit,osv-scanner -r .; skillsupply-chain-risk-auditor.- Lockfile committed, exact versions, no abandoned or typo-squatted packages, no
postinstallsurprises.
Phase 3 — Authentication and sessions
- Use a proven provider (Neon Auth, Auth.js, Clerk, Supabase Auth). Never hand-roll password crypto.
- Passwords: argon2id/bcrypt; rate-limit login, signup, password reset; generic error messages.
- Cookies:
HttpOnly; Secure; SameSite=Lax(orStrict); rotate session on login; real logout. - Reset tokens: single-use, short-lived, random. Optional 2FA for admin.
Phase 4 — Authorization (the #1 real-world bug)
- Every API route / server action checks on the server who the user is and whether they own the record. Hiding a button is not security.
- Test IDOR: user A changes an id in the URL/body to user B's record → must get 403/404.
- Database: least-privilege role for the app; Row-Level Security where available.
Phase 5 — Input, injection and XSS
- Validate every input on the server with a schema (zod/valibot). Reject unknown fields.
- SQL only through parameterized queries / ORM; never string-concatenate.
- No
dangerouslySetInnerHTML/innerHTML/evalwith user data; sanitize rich text (DOMPurify). - File uploads: check type by content, size limit, random names, store outside the app, never execute.
- SSRF: server never fetches arbitrary user-given URLs without an allow-list.
- Skills
semgrep(runsemgrep --config auto) andsharp-edgesfor dangerous APIs.
Phase 6 — Payments (cards, Pix) — critical
- The site never sees or stores card numbers. Use hosted checkout: Stripe Checkout / Payment Links, Mercado Pago Checkout Pro, or the provider's hosted fields. That keeps the site out of most PCI-DSS scope (SAQ A). Never log, save or send card data to your server.
- Price and total are computed on the server from your own database — never trust an amount sent by the browser.
- Order is marked paid only by a webhook whose signature is verified (Stripe
constructEvent, Mercado Pagox-signature), handled idempotently (same event twice = one payment). - Secret keys server-only; test mode vs live mode keys separated; restricted keys where possible.
- Card owner (e.g. the hairdresser) configures payouts in the provider dashboard herself — her bank/card details never go into the code or chat.
Phase 7 — Business logic and bugs
- Can someone book the same slot twice, use a coupon infinitely, get negative quantities, skip a step, replay a request? Race conditions on anything with money or limits.
- Error handling: no stack traces to users; errors logged server-side.
- Skill
systematic-debuggingfor every bug found;differential-reviewon recent changes.
Phase 8 — Headers, transport and configuration
- HTTPS only + HSTS;
Content-Security-Policy(nounsafe-inlinescripts if avoidable),X-Content-Type-Options: nosniff,Referrer-Policy: strict-origin-when-cross-origin,Permissions-Policy,frame-ancestors 'none'(clickjacking). - CORS: explicit origins, never
*with credentials. CSRF protection on cookie-auth forms. - Rate limiting on all public APIs and forms (e.g. Vercel Firewall / Upstash ratelimit); bot protection on contact/booking forms (Turnstile/hCaptcha).
- Debug mode off, source maps not public, admin routes not guessable and protected,
directory listing off,
robots.txtdoesn't reveal secrets.
Phase 9 — Attack it for real (dynamic test, own preview deploy only)
- OWASP ZAP baseline:
docker run -t ghcr.io/zaproxy/zaproxy:stable zap-baseline.py -t <preview-url>(andzap-full-scan.pyon staging). - Playwright (
webapp-testing): try XSS payloads in every field, IDOR swaps, expired sessions, direct calls to protected APIs without a cookie, uploading a.html/.svgwith script. - Check headers:
curl -sI <url>; TLS grade via ssllabs.com; securityheaders.com. - Optional deep test: Shannon (KeygraphHQ/shannon, AI pentester) against staging — needs Docker and an API key; the user runs it.
Phase 10 — Privacy, legal and resilience (LGPD)
- Run the
legal-complianceskill: privacy policy, terms, cookie consent that really blocks scripts until accepted, consumer info, data-deletion channel. - Collect the minimum personal data; privacy policy page; way to delete one's data.
- Database backups on (Neon point-in-time restore / branches); tested restore.
- Error monitoring (Sentry) and uptime alert. Dependency updates (Dependabot/Renovate).
- Games: validate scores/progress on the server if there is a leaderboard or prize — the client can always be modified.
Phase 11 — Fix, prove, repeat
For each finding: fix at the narrowest place (surgical-patch), add a test that would have
caught it, re-run that phase's check, record the proof in SECURITY-AUDIT.md.
Skill fp-check to discard false positives with evidence. Then run phases 1, 2, 5 and 9
again in full. Done only when there are zero open critical/high findings and every
fix has proof (verification-before-completion).
Final report to the user (Portuguese, plain words)
What was checked, what was found, what was fixed, what still needs them (e.g. turn on 2FA in the hosting account, the business owner setting up payouts), and a short checklist to repeat before each future launch.
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.

