The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes several standard developer CLI tools for local code analysis and testing. This includes pytest for running skill tests, pydeps and madge for generating dependency graphs, and standard Unix utilities like grep, find, rg, head, and git to inspect the codebase structure and ADRs.
[INDIRECT_PROMPT_INJECTION]: The skill ingest untrusted data in the form of repository files (source code, ADRs, and git diffs) during the review workflow. However, it implements a robust grounding mechanism by requiring every finding to include a specific file location and verbatim anchor text, which is then validated by an automated citation verifier to ensure the agent's output is based on actual file content.
[SAFE]: The skill's operations are confined to the local repository environment. It relies on internal dependencies from the imbue namespace for structured output and risk assessment. No unauthorized network operations, credential harvesting, or persistence mechanisms were detected. The architectural lenses provided (e.g., Law of Demeter checks, ceremony audits) are standard software engineering practices.

architecture-review