A Claude Code skill bundle for bug hunting and external red-team work - 82 skills, 15 slash commands, 681 disclosed-report patterns curated across 24 core vulnerability classes, plus enterprise identity + infrastructure attack matrices.
Python ai-securityanthropicapplication-securitybug-bountybugbountybugcrowdclaudeclaude-codeclaude-skillsethical-hackinghackeroneoffensive-securitypentestingred-teamsecurity-toolsweb-security
#!/usr/bin/env python3
"""
gen_skill_index.py — build cbh/data/skill_index.json from the repo.
The `cbh` CLI needs two things at runtime: each skill's `description:` frontmatter
(for `cbh classify`'s matcher) and the list of disclosed-report filenames (shown as
pointers). When `cbh` is pip-installed without the full repo, it falls back to this
generated index instead of reading skills/ and docs/disclosed-reports/ directly.
Run after adding/editing skills:
python3 scripts/gen_skill_index.py
"""
from __future__ import annotations
import json
import re
from pathlib import Path
REPO = Path(__file__).resolve().parent.parent
SKILLS = REPO / "skills"
REPORTS = REPO / "docs" / "disclosed-reports"
OUT = REPO / "cbh" / "data" / "skill_index.json"
DESC_RE = re.compile(r"^description:\s*(.+?)(?=\n[a-z_]+:|^---|\Z)", re.M | re.S)
def main() -> int:
skills: dict[str, str] = {}
for d in sorted(SKILLS.iterdir()):
sm = d / "SKILL.md"
if not d.is_dir() or not sm.exists():
continue
m = DESC_RE.search(sm.read_text(encoding="utf-8"))
if m:
desc = m.group(1).strip().strip('"').strip("'").strip()
skills[d.name] = desc[:2000]
reports = sorted(p.name for p in REPORTS.glob("hunt-*.md")) if REPORTS.exists() else []
OUT.parent.mkdir(parents=True, exist_ok=True)
OUT.write_text(json.dumps(
{"skills": skills, "reports": reports}, indent=1, sort_keys=True) + "\n",
encoding="utf-8")
print(f"wrote {OUT.relative_to(REPO)} — {len(skills)} skills, {len(reports)} reports")
return 0
if __name__ == "__main__":
raise SystemExit(main())