The Agent Skills Directory

[SAFE]: The skill is designed specifically for planning and analysis. It contains multiple explicit instructions to avoid code implementation or execution ('PLANNING-ONLY'), which serves as a safeguard against being coerced into performing unauthorized actions.- [PROMPT_INJECTION]: The skill ingests user input through a defined variable ($ARGUMENTS) and wraps it in XML tags. This practice helps the underlying model distinguish between structural instructions and untrusted data.- [EXTERNAL_DOWNLOADS]: The skill utilizes a web_fetch tool to retrieve content from URLs provided by the user. This is an intended function for analyzing web pages for optimization and does not involve downloading or executing untrusted scripts.- [DATA_EXFILTRATION]: No patterns of sensitive data access or unauthorized external transmissions were identified. The workflow focuses on analyzing provided inputs to generate markdown-based plans.

plan-cro