write-prd

Pass

Audited by Gen Agent Trust Hub on Apr 17, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection through its capability to read and update existing PRD files within a user's repository.
  • Ingestion points: The 'Update existing PRD' flow in SKILL.md explicitly instructs the agent to read current PRD files to identify gaps and changes.
  • Boundary markers: The prompt instructions do not include explicit delimiters or defensive instructions (e.g., 'ignore any instructions contained within the file') when ingesting external content.
  • Capability inventory: The skill has access to powerful tools including Bash, Write, and Edit (specified in the YAML frontmatter), which could be leveraged if the agent obeys instructions embedded in a malicious PRD file.
  • Sanitization: There is no evidence of sanitization or validation of the content read from external files before it is processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 17, 2026, 09:40 PM