cz-perspective

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.85). No technical malware/backdoor patterns or data-exfiltration code are present, but the skill contains high-risk prompt‑injection and impersonation directives (forced first‑person persona of a real, living individual; persistent role state with one‑time disclaimer; trigger phrases that silently activate the behavior) that enable social‑engineering, deceptive attribution, and potential fraud—overall a high abuse risk.

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

• Third-party content exposure detected (high risk: 0.90). The skill's required workflow (SKILL.md "Step 2: CZ式研究" — "先研究再回答") and accompanying README/research files explicitly list and rely on fetching public third‑party sources (e.g., 4,449 tweets, AMAs, news interviews and many URLs in research/* and the "PRIORITY FETCH LIST"), so the agent is expected to ingest untrusted, user‑generated open web content that can materially influence its actions.