Security workflow skill for safely reviewing and adopting untrusted external code, skills, packages, and MCP/tool integrations.
A Claude Code skill (SKILL.md) that provides a structured protocol for reviewing external content before it enters your development environment. Apply it whenever you're working with:
- Unknown repos or packages
- Install scripts from the internet
- MCP server descriptions or tool definitions
- Prompt templates from external sources
- Any content that could contain prompt injection
Copy SKILL.md to your Claude Code skills directory:
cp SKILL.md ~/.claude/skills/cyberwoods/SKILL.mdOr include in your project's .claude/skills/ directory.
SKILL.md— The skill definition (trigger rules + protocol)agents/openai.yaml— OpenAI agent format for cross-platform usereferences/threat-model.md— Threat model referencereferences/adoption-checklist.md— Checklist for safe adoption
MIT