render-67js-media
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill references the execution of local utility scripts such as
scripts/render-media.mjsandscripts/score.mjs. These are intended for contributor workflows and operate on local scene data within the repository's root directory. - [INDIRECT_PROMPT_INJECTION]: The agent is instructed to ingest and act upon data from rendering tool outputs, including status messages and JSON metrics (
scene_render_status,scene_render_result). While this introduces untrusted data into the context, the skill mitigates risk by providing clear instructions on polling logic and result verification, and the surface is limited to rendering metadata. - [DATA_EXFILTRATION]: The documentation explicitly prohibits including local file paths when reporting issues to the developer, which serves as a safeguard against the accidental disclosure of sensitive file system information.
Audit Metadata