claude-authenticity

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). These links point to a runnable Python script hosted on a third‑party aggregator (claudskills.com) plus an unfamiliar GitHub repo and placeholder API endpoints — there are no direct .exe downloads, but copying/running unreviewed code that asks for API keys and targets arbitrary endpoints is a moderate–high risk (credential exfiltration or malicious endpoint testing), so treat as suspicious.

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime path calls the user-supplied ENDPOINT with the user-supplied API_KEY, then parses the returned JSON and extracts answer/thinking text from the response (_call → resp.json() → _extract_answer/_extract_thinking), so any free-form text in the provider’s response (outsider-authored) is ingested into the agent’s LLM context for scoring/prompt-injection checks.