prd-review-panel
Pass
Audited by Gen Agent Trust Hub on Apr 2, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of PRD files.
- Ingestion points: The workflow in
SKILL.mdreads content from PRD files located inoutputs/prds/andcontext-library/prds/to be reviewed. - Boundary markers: The sub-agent prompts (Engineering, Design, Executive, etc.) interpolate the PRD text following a simple
PRD Content:header. This lacks strong delimiters (like XML tags or unique random tokens) or explicit instructions to the sub-agents to ignore instructions embedded within the document. - Capability inventory: The skill can read local files, write synthesis reports to the
outputs/prds/directory, and trigger parallel sub-tasks using the platform'sTasktool. - Sanitization: There is no evidence of sanitization, escaping, or validation of the input PRD content before it is passed to the sub-agents, allowing a malicious document to potentially influence the analysis output.
Audit Metadata