Cybersecurity Skills
Cybersecurity Skills for AI Agents
42 security skills for Claude Code, Cursor and Codex: code and cloud audits, authorization-gated pentesting and incident response
- 501 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The offensive skills send requests to targets and can harm systems without the owner's permission
- `allowed-tools` pre-approves the listed tools without a prompt in Claude Code
- Audits read source, Git history and configs, which can hold secrets
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/cybersecurity-skills-for-ai-agentsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The offensive skills send requests to targets and can harm systems without the owner's permission
- `allowed-tools` pre-approves the listed tools without a prompt in Claude Code
- Audits read source, Git history and configs, which can hold secrets
This is third-party code. Review the repository files before installing.
What it does
Cybersecurity Skills is a set of 42 information-security skills in `SKILL.md` format, grouped into seven families: application and supply-chain security, offensive recon, detection and response, cloud and infrastructure, AI security, design and governance, and compliance and privacy. The skills are written so an agent can run them end to end and explain the findings in plain language. The `owasp-audit` skill checks source code against OWASP Top 10:2025: architecture recon first, then a per-category checklist with grep patterns, then a refutation pass on every candidate. The result is `findings.json` with confirmed, needs validation and rejected statuses plus a SARIF export. The same scheme backs `api-audit`, `container-audit` and `crypto-audit`. Among the others, `secrets-audit` hunts for leaked keys in code and Git history, `scanner-pipeline` wires gitleaks, semgrep and trivy into CI, `malware-triage` examines a suspicious file without running it, and `agent-supply-chain-audit` checks the agent's own skills, plugins and MCP servers. The offensive skills (`recon`, `osint-recon`, `web-pentest`, `red-team-engagement`) begin with an authorization check on the target and refuse ambiguous requests. If you do not know where to start, run `security-start`: it asks up to three questions and names the right skill.
Who it is for. Security engineers, developers without a dedicated security team and founders of small teams who need audits of code, cloud and dependencies done through an agent.
Good fit when
- Before a release you want to walk the source against OWASP Top 10:2025 and get a report of confirmed findings
- A loud vulnerability dropped and you run many apps, so you need to learn which are affected with the `patch-sweep` skill
- Logs are collected but there is no SIEM: `log-watch` sets up scheduled review and paging
- Before trusting someone else's skill, plugin or MCP server, `agent-supply-chain-audit` checks it
- You want a pentest of a web app you have written authorization for: `web-pentest` follows nine WSTG phases
Not a fit when
- The target is not yours and you have no written authorization: the offensive skills refuse such requests
- You expect to replace a security specialist: the author says the skills give a first pass its structure and do not replace a specialist
- The agent cannot read the repository or run tools: most audits rely on reading code
- You plan mass scanning of other people's systems or destructive techniques: the skills refuse those
Example request
Run owasp-audit on this repository in standard mode and compile a report of confirmed findingsLimitations
The offensive skills work only with explicit authorization for the target, and the author states plainly that using them against systems you lack permission for is wrong. The Cursor and Codex adapters do without the `finding-verifier` subagent and repeat its steps from the `verification.md` file. The `npx skills` installer takes one skill folder at a time, so shared references are copied inside each. Verify the agent's findings by hand; the `allowed-tools` field in Claude Code pre-approves the listed tools without a prompt and is not a sandbox.
How to disable. Delete the skill folders from `.agents/skills/` and `.claude/skills/`, or run `/plugin uninstall cybersecurity-skills`. Reports and `findings.json` stay where the agent wrote them.
FAQ
How do I install just one skill, such as owasp-audit?
Run `npx skills add briiirussell/cybersecurity-skills --skill owasp-audit`. `npx skills add briiirussell/cybersecurity-skills --list` prints every skill. By hand, copy the `skills/owasp-audit` folder into a project's `.claude/skills/` or into `~/.claude/skills/`.
What is finding-verifier?
A subagent that ships with the Claude Code plugin. The audits hand it one finding at a time, and in a fresh read-only context it tries to disprove the finding. Cursor and Codex get no subagent and follow the same steps from the text of `verification.md`.
How do I tell a confirmed finding from a doubtful one?
In `findings.json` every record has a status of confirmed, needs validation or rejected. The first means the refutation pass found no objection, the second needs a manual check, and the third lists discarded candidates. The same file exports to SARIF.
Which skill fits if I have no security specialist?
Start with `security-start`, and for your own accounts and devices with `personal-opsec`. Use `dependency-audit` for dependencies, `secrets-audit` for leaked keys and `log-watch` for logs without a SIEM.
Is the quality of the skills themselves checked?
Every PR passes a skills linter and a static NVIDIA SkillSpector scan, and tests cover the findings schema, SARIF export and adapter builds. You can rerun the scan with `skillspector scan https://github.com/briiirussell/cybersecurity-skills --recursive --no-llm`.
Discussion
No questions or reviews yet.
Related
SkillsClaude Code CyberSecurity Skills
Claude Code CyberSecurity Skill Collection
A set of 22 cybersecurity skills for Claude Code: recon, vulnerability assessment, incident response, red and blue team, malware analysis
A security scanner for agent skills: finds prompt injection, data exfiltration and malicious code before you install
A code security audit skill by Cloudflare: the agent runs recon, coverage-led hunting and independent verification of findings, then produces a structured repor
A cybersecurity skill library: forensics, incident response, pentesting, cloud and malware analysis mapped to MITRE ATT&CK and NIST
A skill for reviewing code for vulnerabilities against OWASP Top 10:2025, ASVS 5.0 and the 2026 lists for LLM and AI agent security
Security audit skill with GOST and FSTEC
Security Audit Skill for Claude Code
A security audit skill for Claude Code: OWASP Top 10 plus GOST R 56939, FSTEC Order No. 21 and 152-FZ, parallel subagents and a compliance matrix
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Cybersecurity Skills (this card) | no | 501 | |
| Anthropic Cybersecurity Skills | no | 35,628 | a skill library tied to MITRE ATT&CK and NIST, while here every finding passes a refutation pass and gets a status in `findings.json` |
| OWASP Security | no | 429 | a single skill for checking code against OWASP, while here the OWASP audit is one part of a set of 42 skills |
README in short
The description lists audiences from pentesters to founders with no security staff, plus hints on which skill to start with in each situation. It then shows the family map, a table of all skills, seven ways to install, rules for building adapters and writing a new skill, and sections on security scanning and on ethics and authorization.
README badge
Original SKILL.md
--- name: owasp-audit description: "Audit application source code against the OWASP Top 10:2025 categories — broken access control (incl. SSRF), security misconfiguration, software supply chain failures, cryptographic failures, injection, insecure design, authentication failures, software and data integrity, logging and alerting failures, mishandling of exceptional conditions — with recon, a coverage ledger, per-finding verification, and a findings.json report. Use when the user mentions 'OWASP,' 'OWASP Top 10,' 'security audit,' 'security review,' 'secure code review,' 'code security review,' 'vulnerability audit,' 'find vulnerabilities,' 'appsec review,' 'application security audit,' 'check for security issues,' 'review this PR for security,' 'pre-merge security check,' 'diff review,' 'broken access control,' 'IDOR,' 'SQL injection,' 'XSS,' 'SSRF,' 'webhook security,' or wants to check their codebase for common security weaknesses." allowed-tools: Read, Grep, Glob, Bash(npm audit), Bash(npm audit --omit=dev) --- # OWASP Audit — Source Code Security Review Perform a systematic security audit of application source code against the OWASP Top 10:2025. Every candidate is traced, then independently challenged, before it becomes a finding. Coverage is recorded per surface and trust boundary, so "clean" means someone actually read the code. Cross-references: - `api-audit` for a surface-driven pass over every API endpoint (shares the recon, verification, and orchestration references) - `web-pentest` "Combined white-box + black-box mode" when you have both the source and a running instance you're authorized to test - `business-logic-testing` to prove race conditions and logic flaws dynamically (A06) - `dependency-audit` for the full A03 supply-chain picture; `secrets-audit` for leaked credentials - `threat-modeling` for the principal / trust-boundary vocabulary used in recon - `finding-triage` to take a single finding from this report to a defensible disposition ## Authorization and Mode Only audit code the user owns or is authorized to review. Any live check (for example `curl -I` against a deployed origin) needs the same authorization for that host and stays non-destructive. Pick the mode before starting: - **Guidance** (default): a question, a single file, or "how do I secure X". Answer directly. Findings still follow the verdict rules below. No ledger required. - **Full audit**: the user explicitly asks to audit or security-review a codebase, a release, or a PR. Run the full workflow below. - **Diff / PR-scoped**: the user wants a branch, PR, or commit range reviewed, or a pre-merge gate. Run the full workflow limited to the change, using [references/diff-review.md](references/diff-review.md). If it's ambiguous, ask one question: guidance pass or full audit (and which profile)? ## Workflow (full audit) 1. **Scope**: what's in, what's out, mode, and run profile 2. **Recon**: write `architecture.md` and `coverage.json` before hunting