skill-auditing

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The audit instructs the agent to extract and report concrete commands, API calls and config keys from skill files and to "report each violation" with file and line, which can require copying secret values (API keys, tokens, passwords) verbatim into the report and thus enables exfiltration; the prompt gives no instruction to redact or avoid printing secrets.

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

• Third-party content exposure detected (high risk: 0.75). At runtime the skill’s “Content layer (live)” explicitly performs web search / doc fetch to verify factual claims, which can ingest outsider-authored free text from public web pages into the agent’s LLM context.