The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a helper script (scripts/naming-ledger.py) that executes local git commands using subprocess.run. The script uses structured argument lists rather than shell execution, minimizing the risk of command injection when interacting with the local repository environment.
[PROMPT_INJECTION]: The instructions establish a strict safety protocol for 'Exclusive Execution.' The agent is required to present a plan and wait for an explicit affirmative response from the user before performing any edits or path moves. This prevents the agent from performing unintended or autonomous destructive actions without human oversight.
[DATA_EXPOSURE]: The skill maintains a 'Coverage Ledger' in JSON format to track every path in the repository. The instructions explicitly mandate that this ledger be created outside the repository to prevent the internal state of the refactor or temporary metadata from being accidentally committed or exposed.

naming-refactor