Security
7 skills built from studying how Anthropic designs agents internally. ~700 lines, zero fluff.
npx -y skills add jessedegans/jstack --skill securityAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Use when reviewing code for security vulnerabilities, before shipping to production, or when the user asks for a security audit. Runs OWASP Top 10 + STRIDE analysis with a strict false positive filter. Confidence gate at 8/10. No theoretical hand-waving.
SKILL.md
3.1 KB, 676 tokens by cl100k_base, as published. Nobody here has run it
Security Audit
The Scan
1. Map the attack surface
- Grep for routes, auth handlers, external integrations, file uploads, admin endpoints
- List all entry points (API routes, webhooks, scheduled jobs, CLI commands)
2. OWASP Top 10 pass
For each entry point, check:
| Category | Look for |
|---|---|
| A01 Broken Access Control | Missing auth checks, direct object references, path traversal |
| A02 Crypto Failures | Weak algorithms, hardcoded secrets, plaintext storage |
| A03 Injection | SQL, command, template, LLM prompt injection |
| A04 Insecure Design | Business logic flaws, missing rate limits |
| A05 Misconfiguration | Debug mode, default creds, overly permissive CORS |
| A06 Vulnerable Components | Outdated deps with known CVEs |
| A07 Auth Failures | Weak passwords, missing MFA, session fixation |
| A08 Integrity Failures | Unsigned updates, untrusted CI/CD inputs |
| A09 Logging Gaps | Missing audit trail, logging sensitive data |
| A10 SSRF | User-controlled URLs fetched server-side |
3. STRIDE per component
For each significant component: can an attacker achieve Spoofing, Tampering, Repudiation, Information Disclosure, Denial of Service, or Elevation of Privilege?
4. Data classification
Tag all stored/transmitted data as:
- RESTRICTED - credentials, PII, financial data
- CONFIDENTIAL - internal business logic, configs
- INTERNAL - non-sensitive but not public
- PUBLIC - safe to expose
The Filter (the actually valuable part)
Hard exclusions, don't report these:
- DoS via resource exhaustion (unless trivially exploitable)
- Test-only code vulnerabilities
- Memory safety in memory-safe languages
- Log spoofing / log injection
- Missing CSP headers (unless XSS is also present)
- Theoretical timing attacks without proof
- Missing X-Frame-Options when CSP frame-ancestors is set
- Self-XSS (requires attacker to control own browser)
- Open redirects without auth token leakage
- Missing rate limiting on non-auth endpoints
- Verbose error messages in development mode
- Cookie without Secure flag on localhost
- Missing HSTS on internal-only services
- Theoretical ReDoS without proof of exploitable input
- Information disclosure via stack traces in dev mode
Precedents, assume these are fine unless proven otherwise:
- UUIDs are unguessable
- Environment variables are trusted
- Client-side JS doesn't need server-side auth
- ORM-generated queries are injection-safe
- Framework CSRF protection works when enabled
- HTTPS is enforced at infra level
- Container isolation is effective
- Cloud IAM roles are correctly scoped
Rules
- Confidence gate: 8/10 minimum. If you're not at least 80% sure it's real, don't report it.
- Every finding needs an exploit scenario. Step-by-step: what does the attacker do, what do they get? No hand-waving.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.