ensemble-content-scorer
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
- [PROMPT_INJECTION] (LOW): The skill is susceptible to indirect prompt injection because it processes untrusted user-provided content ideas through multiple external LLM APIs. Ingestion points: User input via triggers like 'Ensemble score: [topic]' or CLI arguments such as '--idea'. Boundary markers: No markers, delimiters, or isolation instructions are documented to distinguish user data from the agent's instructions. Capability inventory: The skill uses the 'anthropic', 'openai', and 'google-generativeai' libraries to send data to third-party AI models. Sanitization: No sanitization, validation, or escaping of the user-provided content ideas is documented.
- [NO_CODE] (SAFE): The skill's documentation and usage examples rely on a script located at 'scripts/score_content.py', which is missing from the provided file set. While this is a best practice violation and renders the skill non-functional in its current state, it does not present a direct security threat.
Audit Metadata