bmad-generate-project-context
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill forwards untrusted user data (architecture docs, specifications, and preferences) verbatim to a downstream skill.
- Ingestion points: User-supplied requests and files containing project details are processed in the
SKILL.mdshim. - Boundary markers: None present; the skill is explicitly instructed to forward inputs "verbatim" without delimiters or warnings to ignore potentially embedded instructions.
- Capability inventory: This specific shim script does not contain system-level capabilities such as direct file writing or network operations; however, it hands off control to the
bmad-project-contextskill which performs those actions. - Sanitization: None; input is passed through directly as provided by the user without validation or filtering.
- [EXTERNAL_DOWNLOADS]: The manifest file references a remote repository used for module updates.
- Evidence: The
update_sourcefield inmodule-manifest.mdpoints togithub:bmad-code-org/bmad-skills/skills. - Context: This is a standard reference to the vendor's own repository for skill maintenance and updates.
Audit Metadata