continuous-learning-v2

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). The URL list includes an embedded-credentials GitHub URL (https://ghp_xxxx@github.com/) — a high-risk indicator (exposed token/credential) that can enable unauthorized access or distribution of malicious code; the other links are ordinary Git/GitHub/example/X references and not direct, obviously-malicious download hosts.

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

• Third-party content exposure detected (medium risk: 0.30). The observer background agent (agents/observer-loop.sh + agents/observer.md) ingests tool input/output strings recorded by hooks/observe.sh into local observations.jsonl, and those strings originate from arbitrary user/tool text during Claude Code sessions (outsider-authored content can be posted into the session and thus stored/LLM-analyzed).