The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local shell commands to gather architectural data and metrics from the repository.
Evidence: git rev-parse --show-toplevel is used to identify the root of the project.
Evidence: ripgrep (rg) is employed by multiple subagents to search for layering violations, type strictness issues, and code smells.
Evidence: wc -l is used to evaluate file and function sizes for hygiene reporting.
[PROMPT_INJECTION]: The skill processes untrusted repository content, creating an attack surface for indirect prompt injection.
Ingestion points: Subagents ingest all source code and test files in the src/ and tests/ directories as part of the audit process.
Boundary markers: There are no explicit delimiters or instructions provided to the subagents to disregard potential instructions embedded within the source code being analyzed.
Capability inventory: The skill can spawn new subagents via the Agent tool and execute local shell commands for data processing.
Sanitization: The skill does not implement sanitization or escaping of the repository content before passing it to the subagents for evaluation.

audit-architecture