bmad-os-diataxis
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided documentation files to perform updates and refinements using subagents. This process creates a surface for indirect prompt injection, as malicious instructions hidden within the target documentation could potentially manipulate the agent's behavior during the editing process.
- Ingestion points: Documentation files targeted for creation, update, or refinement (e.g., files in
docs/). - Boundary markers: The subagent prompt for refinement tasks lacks explicit delimiters (such as triple backticks or custom markers) or specific instructions to ignore any commands found within the target document content.
- Capability inventory: The skill makes extensive use of the
Edit toolto modify the local file system. - Sanitization: No sanitization, escaping, or filtering of the input file content is performed before the content is passed to the subagents for processing.
Audit Metadata