The Agent Skills Directory

[COMMAND_EXECUTION]: The skill identifies the project technology stack by reading configuration files such as package.json, go.mod, and Cargo.toml. This file access is limited to metadata discovery for prompt context and does not involve executing shell commands or modifying files.
[PROMPT_INJECTION]: The skill processes untrusted user-provided draft prompts to generate optimized versions. It mitigates potential injection risks by enforcing a strict advisory-only constraint and explicitly prohibiting the execution of any tasks or instructions contained within the user input.
[SAFE]: The skill includes robust guardrails that prevent it from switching into implementation mode, ensuring it only provides analysis and optimized prompt text rather than performing actions.

prompt-optimizer