gds-prd
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a Python script located at
{project-root}/_bmad/scripts/resolve_customization.pyusinguv runduring the activation phase to resolve workflow settings.\n- [COMMAND_EXECUTION]: The skill uses system-level commands (open,xdg-open, orstart) to automatically display generated HTML validation reports in the user's default browser.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, such as Game Design Documents (GDD) and research materials, to populate the PRD draft. While it utilizes a subagent extraction pattern to mitigate direct exposure, this processing remains a potential vector for indirect prompt injection.\n - Ingestion points: User-provided documents (GDD, research), file globs specified in
persistent_facts, and data fromexternal_sources.\n - Boundary markers: The instructions mandate an "Extract, don't ingest" approach, delegating processing to subagents to avoid loading untrusted content into the primary agent context.\n
- Capability inventory: The skill possesses capabilities for file system writes (
prd.md,decision-log.md), system command execution (uv run,open), and integration with external tools through handoff directives.\n - Sanitization: The skill relies on structural extraction by subagents rather than explicit sanitization or escaping of the ingested text.
Audit Metadata