Finds and fixes vulnerabilities in code the user owns across web, API, and LLM surfaces, and owns confidentiality for authorized remote credential handoffs. Use for a security review, "is this safe to ship," "check for vulnerabilities," "보안 점검," secrets hygiene, dependency risk, PR security regressions, or "hand a secret to a remote agent." Not for building LLM-agent systems, Orca session operation (use official `orca-cli`), or general hook and gate configuration; this skill never attacks.
Create security architecture diagrams using PlantUML syntax with identity, encryption, firewall, and compliance stencil icons. Best for IAM flows, zero-trust models, encryption pipelines, and threat detection architectures.
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.
Choose the next security skill for a new repository, a weekly sweep, or a leaked credential. Explains priorities, safeguards before each fix, and how to measure its result.
Use when you want a dedicated security review of staged or recently changed files — deeper than the security section in /review. Covers OWASP Top 10, secrets exposure, dependency vulnerabilities, and auth gaps. Run before opening a PR on security-sensitive changes.
Use when auditing code, dependencies, or infrastructure for security issues (SAST, dependency scanning, secret leak detection, OWASP-style checks), or when checking that a project or change follows secure defaults before committing or deploying.
Provides expert knowledge for implementing and reviewing application security following OWASP Top 10 (2021), with stack-specific patterns for Spring Boot, Django, Next.js, and Astro.
Provides expert knowledge for implementing and reviewing application security following OWASP Top 10 (2021), with stack-specific patterns for Spring Boot, Django, Next.js, and Astro.
Project-wide security audit (OWASP Top 10, auth, access control, headers, secrets, dependencies). Run before go-live and after major changes. Mandatory gate before /deploy prod.
Investigate security findings within an authorized scope, with bounded read-only investigation and explicit coverage and evidence limits. Covers code hunts and product-level review (standard coverage, harm to people, privacy, detection).
The single security skill — diff/PR audit method with a high-confidence bar (BOLA/IDOR, mass assignment, DTO validation, injection, secrets, rate limiting), STRIDE threat modeling before large features, backend auth hardening (guards, Passport, RBAC, CORS/CSRF, OWASP ASVS), and mobile hardening (MASVS: token storage, deep links, WebView, build config). Load the SKILL.md for routing, then read ONLY the references/ file matching the job.
Audits and hardens code against the OWASP Top 10 — input validation, authn/authz, secrets management, cryptography, and dependency scanning with CWE mapping. Trigger with "security review", "is this endpoint secure", "scan our dependencies", or "where do we store this secret".
Apply security best practices to protect applications, data, and users from vulnerabilities. TRIGGER this skill automatically when the user mentions: SQL injection, XSS, CSRF, authentication bypass, input validation, sanitizing user input, environment variables or secrets management, HTTPS, rate limiting, permission checks, authorization, security audit, vulnerability, or any phrase like "is this secure?", "how do I protect", "prevent attacks", or "secure this endpoint". Also trigger when reviewing code that handles user-supplied data, passwords, tokens, API keys, or database queries.
Plain text files in a repository that tell a coding agent how the project works: commands to run, conventions to follow and things to avoid. CLAUDE.md, AGENTS.md, cursor rules and skills are the common kinds.
CLAUDE.md or AGENTS.md?
CLAUDE.md is read by Claude Code. AGENTS.md is an open format that Codex, Cursor and other agents read. Many projects keep one and point the other at it.
What is a skill?
A folder with a SKILL.md that describes one capability, such as filling PDFs or reviewing code. The agent loads it only when the task calls for it.
Can I search my own team's files too?
Your agents already can, over MCP, limited to the files you're allowed to read. Searching them from this page is coming.