The Agent Skills Directory

[SAFE]: The skill defines a narrow, behavior-preserving contract and explicitly refuses to perform security reviews or debt analysis. No malicious patterns or unauthorized network/file access were identified in the instructions or references.- [COMMAND_EXECUTION]: The skill implements shell hooks to execute git diff before and after file modifications. These are benign development practices used to warn about uncommitted changes and provide a summary of the edits performed by the agent.- [PROMPT_INJECTION]: The skill is a surface for indirect prompt injection because it processes user-provided code targets and snippets. This risk is managed by instructions requiring explicit verification of behavior preservation and invariant reasoning to ensure embedded data does not influence the agent's logic.

simplify