skill-audit

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 1.00). These URLs point to untrusted external endpoints (example/unknown/“evil” domains) used in the skill examples to receive POSTed data — a classic data-exfiltration/webhook pattern and high-risk indicator for malware or privacy abuse.

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

• Third-party content exposure detected (medium risk: 0.30). Runtime path reads free text from the skill/agent-definition directory provided at audit time (e.g., SKILL.md and all files under the skill directory via scan_skill()/collect_text_files()), so an outsider can supply that content for the LLM to ingest for analysis.