pitfall-skill

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). Runtime path: the agent workflow writes the outsider-authored message text the user provides into .pitfalls/ledger.yaml / .pitfalls/entries/*.md via the log step (e.g., scripts/cli.js:77-159), and later resolve/promote reads that free text back and emits it into CLAUDE.md/AGENTS.md/*.mdc for the LLM to use.