The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a Python script (scripts/naming-ledger.py) to execute standard Git commands such as ls-files, status, and rev-parse. These operations are used to maintain a deterministic ledger of file paths during refactoring. The subprocess calls are implemented using list-based arguments without shell invocation, following security best practices to prevent command injection.
[DATA_EXPOSURE]: The skill generates a ledger file to track the refactoring progress. The helper script explicitly enforces that this ledger must be stored outside the repository directory, preventing the accidental exposure or commitment of metadata to the source control.
[INDIRECT_PROMPT_INJECTION]: The skill processes file paths and contents within a repository to derive naming conventions. This ingestion of untrusted data represents a theoretical surface for indirect prompt injection. However, the risk is mitigated by the skill's structured approach, which includes verification waves and explicit coordination steps.

naming-refactor