action-continuity-choreography-by-d5g808
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from a local reference file (
references/cases.md) which contains prompt excerpts and links, creating a potential attack surface. \n - Ingestion points: The agent is instructed to read
references/cases.mdon every invocation to select a reference anchor case. \n - Boundary markers: The skill contains instructions to distinguish observed evidence from the agent's own recommendations, but lacks explicit structural delimiters or 'ignore' instructions for the ingested text. \n
- Capability inventory: The agent is directed to use a browser tool to inspect external media URLs (mp4, jpg) provided in the evidence file. \n
- Sanitization: No explicit sanitization, validation, or escaping of the retrieved content is mentioned before it is processed by the agent. \n- [EXTERNAL_DOWNLOADS]: The skill contains references to external media and case documentation hosted on
goodcase.aiandx.com. \n - Evidence: The file
references/cases.mdcontains multiple links to.mp4and.jpgfiles (e.g.,media.goodcase.ai/cases/2623f3a18f10.mp4) as well as social media posts.
Audit Metadata