Contributing
May 5, 2026 · View on GitHub
PRs and issues welcome.
What's in scope for contributions
- OOS rebuttal templates for additional clauses — programs use varied OOS language across H1, Bugcrowd, Intigriti, Immunefi
- Per-class hunt skills focused on niches (fintech-specific, healthcare FHIR, government compliance bugs)
- Improvements to
huntshell scaffold — alternative folder layouts, additional file templates, integrations with secret managers - VRT mapping additions for finding types not yet in the table in
bugcrowd-reporting/SKILL.md - Evidence-hygiene additions — new redaction patterns, new tools, new file formats (PCAP, etc.)
- Documentation improvements — clearer USAGE.md sections, more worked examples, better INSTALL.md troubleshooting
What's NOT in scope
- Substantive modifications to vendored upstream skills — submit those upstream (e.g. to shuvonsec/claude-bug-bounty) and we'll pull them in on the next refresh. Path-consistency tweaks are fine.
- Skills that include actual exploitation payloads against specific targets — keep things abstract / class-based
- Personally-identifiable bug-hunting engagement data — anonymize all examples (target names, account UIDs, endpoint names, bounty amounts)
- Anything that requires non-MIT-licensed dependencies
How to propose a change
Small fix or doc improvement
- Fork the repo
- Make the change on a branch
- Open a PR with a short description of what changed and why
New skill
- Open an issue first describing the skill: name, purpose, what gap it fills
- Get feedback on whether the skill is in-scope before writing the full SKILL.md
- Once agreed, submit a PR with:
skills/<name>/SKILL.md— frontmatter description ≤ 1024 chars, body ≤ 500 lines- Mention in
README.mdandUSAGE.mddecision tree - One worked example showing the skill triggering correctly
Skill quality standards
- Frontmatter:
name(lowercase-hyphen-only) +description(≤ 1024 chars, weaves in trigger keywords as natural prose) - Body: target ~1,500–2,000 words, max ~500 lines
- Detail content: if more is needed, use
references/subfolder pattern (seeoffensive-osint/) - Single responsibility: one skill should do one thing well
- Cross-reference complementary skills: mention which skills compose with this one
- Real examples: prefer worked examples over abstract descriptions
Testing changes
There's no automated test suite. Manual smoke tests:
- Install the skill locally (copy to
~/.claude/skills/) - Open a fresh
claudesession - Ask a question that should trigger the skill (using the keywords from the description)
- Verify Claude triggers the skill and uses its content correctly
- For
hunt-*skills, also verify the auto-trigger works without explicitly naming the skill
Code of conduct
- Be respectful in PRs and issues
- Don't include personal data, credentials, or specific target identifiers in examples
- Cite sources when adapting content from disclosed bug-bounty reports or other community work
Licensing
By contributing, you agree your changes are licensed under the same MIT license as the rest of the repo.