sharing-results
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
watch-skillcommand via theBashtool to generate HTML viewer pages and extract bug reports. This is a core function of the skill using a vendor-specific CLI tool. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data such as video frames, transcripts, and OCR text which are then included in the generated HTML report.
- Ingestion points: Video source, transcripts, and OCR data identified in the
SKILL.mddescription and commands. - Boundary markers: No specific delimiters or instruction-ignore markers are specified for the processed video data.
- Capability inventory:
Bashtool is used for executing thewatch-skillreport generation commands (found inSKILL.md). - Sanitization: The skill does not explicitly mention sanitization of OCR or transcript data before inlining it into the HTML viewer.
Audit Metadata