The Agent Skills Directory

[SAFE]: The skill defines a set of instructions for the agent to follow to perform code cleanup and reduction during development tasks. No malicious behavior or attack vectors were detected.- [COMMAND_EXECUTION]: The skill instructs the agent to use grep to verify the presence of callers before deleting symbols. This is a standard and safe use of development tools within the local repository environment.- [DATA_EXPOSURE]: No sensitive file paths, hardcoded credentials, or network exfiltration patterns were identified. The skill operates locally on the project's source code.- [PROMPT_INJECTION]: While the skill uses assertive language to define agent behavior (e.g., 'Do not ask first', 'the turn is NOT done'), these instructions are scoped to the user-invoked task and do not attempt to bypass core system safety filters or ignore prior security constraints.

less-but-better