free-model-triage
Pass
Audited by Gen Agent Trust Hub on Jul 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns or security risks were detected in the skill.
- [EXTERNAL_DOWNLOADS]: The skill references OpenRouter (
openrouter.ai), which is a well-known service for accessing various LLM providers. These references are part of the intended functionality for routing triage requests. - [DATA_EXFILTRATION]: The skill involves sending data to an external API (OpenRouter). It includes a dedicated 'Guardrails' section that explicitly warns users against sending sensitive or private data (health, legal, client email) to free models due to potentially less restrictive data retention/training policies.
- [COMMAND_EXECUTION]: A bash block is included to verify the configuration. It uses standard utilities (
cat,node) to create a local JSON file and validate its schema. This is a local verification step and does not involve executing code from untrusted sources. - [PROMPT_INJECTION]: The skill's primary function is to process untrusted data like emails and articles for triage. While this creates an indirect prompt injection surface, the skill is limited to generating summaries and flags, and does not grant the agent capabilities that could be abused by malicious content in the processed items.
Audit Metadata