clawdefender

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). Most entries are benign/placeholder domains (example.com, api.example.com) or a well-known platform (github.com), but the list includes webhook.site (an exfiltration/collection endpoint) and the cloud metadata endpoint 169.254.169.254 (SSRF/sensitive-data risk), so the set contains high-risk items and should not be treated as a trusted download source.

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly ingests and processes untrusted third‑party content (e.g., SKILL.md Quick Start and HEARTBEAT.md usage showing "curl -s ... | ./scripts/sanitize.sh --json", "gog gmail read | sanitize.sh", and examples in scripts/sanitize.sh) — sanitize.sh/ClawDefender are designed to read API responses, emails, Trello/GitHub content from the public web and then output/present that content for downstream use, so those external, user-generated sources can materially influence agent behavior.