auto-loom-proof
Pass
Audited by Gen Agent Trust Hub on Jul 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute the project's full test suite to capture results for the video.
- [COMMAND_EXECUTION]: It executes local JavaScript utility scripts (
doctor.mjs,validate.mjs,run.mjs) which are part of the 'auto-loom' engine provided by the vendor. - [COMMAND_EXECUTION]: The skill directs the agent to start a local static web server to host generated proof assets.
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface (Category 8). The skill ingests untrusted data from
factory/HANDOFF.md(acceptance criteria) and uses it to drive browser actions and generate narration scripts. Ingestion points:factory/HANDOFF.mdandfactory/REVIEW.md. Boundary markers: Absent; there are no instructions to the agent to ignore instructions potentially embedded in the requirement criteria. Capability inventory: Executes local test suites, runs browser automation via Playwright, and initiates local network services. Sanitization: Absent; no validation or escaping of criteria text is performed before interpolation into automated tasks.
Audit Metadata