muze-quote-generator
Pass
Audited by Gen Agent Trust Hub on Feb 23, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: No security issues such as malicious command execution, data exfiltration, or obfuscated payloads were identified during the analysis.
- [PROMPT_INJECTION]: The skill processes project data from an internal API to generate quotes, which creates a surface for indirect prompt injection. However, this is considered low risk due to the absence of dangerous tools.
- Ingestion points: External data is ingested from the
/api/projects/{projectId}endpoint during Phase 1. - Boundary markers: The template defined in Phase 2 does not include specific delimiters or warnings to ignore instructions embedded within the fetched project data.
- Capability inventory: The skill's operations are restricted to text template population and submission to an internal
/api/quotesendpoint. It lacks any capability to execute shell commands, perform external network requests, or modify the file system. - Sanitization: No explicit sanitization or validation logic is provided for the data retrieved from the project API.
- [DATA_EXFILTRATION]: The skill interacts with relative API paths (
/api/projects,/api/quotes) belonging to the Muze OS platform. This behavior is consistent with the skill's primary function and does not involve unauthorized external data transfers.
Audit Metadata