bmad-retrospective
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill is designed to execute a local Python script located within the project's internal directory structure (
_bmad/scripts/render_skill.py). This is a standard functional component for developer tools operating on local project data. - [DYNAMIC_EXECUTION]: The skill includes a specialized utility,
scripts/git_evidence.py, which uses thesubprocessmodule to perform git repository analysis. The script implements security best practices by validating the revision range parameter format to prevent command injection and strippingGIT_environment variables to ensure a clean execution context. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the local repository, including git commit messages and markdown-based ticket documentation. To mitigate risks of malicious instructions embedded in these files, the workflow enforces strict provenance rules.
- Ingestion points: Commit subjects, epic/ticket markdown files, and planning documents.
- Boundary markers: The instructions require findings to be based on "deterministic derivation" and explicitly tell the agent to drop any finding that cannot be tied to a verified source reference.
- Capability inventory: The skill possesses capabilities for local file system read/write operations and running git commands via subprocess.
- Sanitization: Input validation is applied to arguments used in git commands to ensure they match expected revision patterns.
Audit Metadata