prd-writer

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: No instructions attempting to bypass safety filters or override agent behavior were detected. The skill focuses entirely on its product management workflow.
  • [EXTERNAL_DOWNLOADS]: The skill does not perform any external downloads or fetch remote code. It relies on a local reference file (references/prd-template.md) for its output structure.
  • [CREDENTIALS_UNSAFE]: No hardcoded API keys, tokens, or credentials were found. There are no patterns suggesting access to sensitive system files.
  • [DATA_EXFILTRATION]: No network-enabled tools or commands (such as curl or wget) are utilized to send data to external domains.
  • [COMMAND_EXECUTION]: The skill does not attempt to execute shell commands, manage processes, or modify the underlying system environment.
  • [SAFE]: The implementation follows best practices for structured agent interactions, including scope freezing and phase-based confirmations, which mitigate the risk of accidental or adversarial deviations.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 11:15 AM
Security Audit — agent-trust-hub — prd-writer