AI skill

What it enables

• “Generate a threat model from our docker-compose.yml” → the assistant runs wyrm init, reviews the baseline, and adds asset sensitivity.
• “wyrm flags WYRM-T001 on the login flow — fix it” → it explains the risk and edits the model (enforce + tag TLS).
• “Add a Cloudflare WAF mitigation” → it adds a mitigation and wires it in.
• “Is this model CI-ready?” → it runs wyrm analyze --fail-on high and resolves what's left.

The skill knows the CLI, the OTM structure, and every STRIDE rule (and how to resolve each one) — the same knowledge the engine enforces.

Install

Claude Agent Skill

Drop the skill folder where Claude looks for skills — the frontmatter description decides when it applies.

mkdir -p ~/.claude/skills/wyrm-threat-modeling
cp skill/SKILL.md ~/.claude/skills/wyrm-threat-modeling/

In claude.ai, paste SKILL.md into a Project's custom instructions.

GitHub Copilot repo instructions

mkdir -p .github
cp skill/copilot-instructions.md .github/copilot-instructions.md

Copilot Chat in that repo now understands wyrm. Or paste the file into a chat.

ChatGPT Custom GPT

Create a Custom GPT (Explore GPTs → Create) and paste skill/chatgpt-instructions.md into its Instructions. Attach SKILL.md and DETECTION.md as Knowledge for deeper answers.

Where it lives

The skill is in the repo under skill/ — one shared body of knowledge packaged for each assistant, with a README of install steps. It's versioned with the tool, so the AI's knowledge tracks the engine.