Instructions for Claude when contributing to this repository.
A library of Claude Code skills for smart contract security, built by CD Security. Each skill is a self-contained capability in its own directory.
audit-prep/ # Solidity audit preparation (Foundry/Hardhat)
SKILL.md # Orchestrator
VERSION # Skill version
references/ # Agent instructions and shared rules
agents/ # Per-agent instructions (testing, source, infra)
shared-rules.md # Common output format for agents
evals/ # Test cases and grading scripts
rust-audit-prep/ # Rust/Solana audit preparation (Anchor/native)
SKILL.md # Orchestrator
references/ # Agent instructions, checklist, shared rules
agents/ # Per-agent instructions (testing, source, infra)
checklist.md # Full check reference
shared-rules.md # Common output format for agents
report-template.md # Detailed report template
- One skill, one purpose.
- Keep SKILL.md under 500 lines — use references/ for agent-specific instructions.
- No gas optimization checks — out of scope for audit preparation.
- No vulnerability analysis — agents must not perform security assessments.
- Every FAIL finding must include a specific, actionable fix.
- Do not commit secrets, API keys, or personal data.
- Test changes against both eval projects (Hardhat + Foundry) before submitting.