Claude Code / Claude Agent SDK skill for authorized offensive security workflows — recon, scraping, stealth automation, hidden-API discovery.
A curated knowledge pack for AI agents performing authorized security work: pentest engagements, bug bounty, CTF, internal red team. Distills 29 OSS tools, four operational playbooks, and a runnable reference site into a single installable skill.
SKILL.md— entrypoint loaded by the agentplaybooks/— runnable workflowsrecon-pipeline.md— ProjectDiscovery suite (subfinder → httpx → katana → nuclei)stealth-automation.md— pydoll vs Camoufox vs Playwright matrixhidden-api-discovery.md— find and replay internal SPA APIs from scratchincremental-sync.md— once an endpoint is mapped, turn it into a sync cursor
references/arsenal.html— interactive field guide to 29 OSS tools (open in browser)scripts/— bash templates ready to copy/pastedocs/— ADRs and longer-form notes
cd <your-project>
mkdir -p .claude/skills
git clone https://github.com/biliboss/skill-sec .claude/skills/secClaude Code will discover SKILL.md automatically. Invoke with /sec or wait for trigger keywords (recon, stealth, hidden api, etc).
git clone https://github.com/biliboss/skill-sec ~/.claude/skills/secOpen references/arsenal.html in a browser. No setup needed.
Use only on systems you own or have written authorization to test. Examples of legitimate use:
- Penetration testing engagement (signed SOW)
- Bug bounty program (in-scope assets per policy)
- CTF / lab environment / your own infrastructure
- Internal red team exercise with management approval
This skill refuses to operate without authorization context. Operators are responsible for compliance with local law (CFAA, GDPR, LGPD, etc).
- Not a "hack any site" toolkit. Every workflow assumes authorized scope.
- Not a replacement for a security professional's judgement.
- Not novel exploits — wraps well-known public tools and methodology.
MIT. See LICENSE.
Active. Knowledge captured from real engagements, sanitized into generic playbooks. PRs welcome.