add-model-descriptions
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS] (SAFE): Fetches model list from
https://router.huggingface.co/v1/models. This domain is associated with HuggingFace, a trusted entity, which downgrades the download finding to SAFE/INFO per [TRUST-SCOPE-RULE]. - [PROMPT_INJECTION] (LOW): The skill is vulnerable to Indirect Prompt Injection as it processes untrusted data from an external API and web search results to modify configuration files.
- Ingestion points: Model IDs are fetched via
WebFetchand model details are gathered via web search inSKILL.md(Workflow steps 1 and 4). - Boundary markers: Absent. There are no instructions for the agent to ignore or delimit potentially malicious content within the fetched data.
- Capability inventory: The skill has file-write access to
chart/env/prod.yamlandchart/env/dev.yaml, and can executegitcommands (Workflow steps 6 and 7). - Sanitization: Absent. The agent interprets external search results and formats them into JSON/YAML without validation or sanitization of the input strings.
Audit Metadata