friction-log

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly requires writing the user's initial request verbatim and including any provided input/codebase in the log (and then cat-ing the log to stdout), which forces the LLM to output any secrets/API keys present in the prompt or input verbatim.

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

• Third-party content exposure detected (medium risk: 0.30). In the required friction-log runtime workflow, the agent reads and records the user’s first-message prompt verbatim and “reads any URLs in the prompt first” (so outsider-authored text can be ingested via user-supplied URL content/content at those URLs) before starting work.