muapi-rednote-cover
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill accepts user-provided text for 'topic', 'style', and 'text_overlay' and interpolates these directly into the image generation instructions without sanitization or boundary markers.
- Ingestion points: Input fields defined in
SKILL.md(topic, style, text_overlay). - Boundary markers: Absent. The skill does not use delimiters to distinguish user-provided data from core instructions.
- Capability inventory: The agent uses the
muapiCLI tool andcurlto interact with external image generation models. - Sanitization: No input validation or filtering is applied to the user strings.
- [DATA_EXFILTRATION]: The skill instructs the agent to make network requests to an external domain that is not included in the standard whitelist.
- Evidence: Guidance to use
curlto interact withhttps://api.muapi.aiwhen CLI aliases are unavailable.
Audit Metadata