bmad-agent-git-archeologist
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill makes extensive use of local shell commands via
gitandgh(GitHub CLI) to search history, view diffs, and inspect PR comments. It also relies on a customhindsightCLI tool for memory storage and retrieval. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external sources including git commit messages, PR bodies, and specifically GitHub PR/review comments (
gh pr view --comments). - Ingestion points:
git log,gh pr view,gh api .../comments, and local documentation files. - Boundary markers: None detected; the instructions do not specify using delimiters or warnings when processing these external strings.
- Capability inventory: The skill has permissions to write to the file system (
_bmad/memory/), create/comment on Plane tickets, and post comments on GitHub PRs. - Sanitization: No explicit sanitization logic or instruction to ignore embedded commands in the source data is present.
- [DATA_EXFILTRATION]: While not performing malicious exfiltration, the skill is explicitly authorized to write project context and 'lost features' to external platforms like Plane (ticket tracker) and GitHub (PR comments), which could involve sensitive project metadata if the agent is misled by the content it analyzes.
Audit Metadata