The Agent Skills Directory

[SAFE]: The skill is purely instructional and designed for text transformation. It uses the disable-model-invocation: true flag in the YAML frontmatter, which ensures the agent does not attempt to call external tools or models during its execution.
[PROMPT_INJECTION]: The skill processes user input via the $ARGUMENTS variable to analyze draft prompts. The instructions explicitly direct the agent to analyze, score, and rewrite the input according to strict templates, which mitigates the risk of the input being interpreted as commands to the agent itself.
[DATA_EXFILTRATION]: No network operations or sensitive file access patterns were detected. While the skill suggests using @ file references in the generated prompts, it does not attempt to read or exfiltrate the contents of those files.
[NO_CODE]: The skill does not provide or download any executable scripts, binaries, or third-party dependencies. It functions entirely through natural language instructions.

prompt-improver