The Agent Skills Directory

[SAFE]: The skill is entirely composed of instructional markdown designed to guide the agent through task planning. It does not contain any executable code, binary components, or external network requests.
[COMMAND_EXECUTION]: The instructions suggest running verification scripts like security_scan.py or api_validator.py. It includes a specific security boundary, directing the agent to only use existing and inspected scripts already present in the project environment, which prevents the execution of unvetted or malicious external commands.
[INDIRECT_PROMPT_INJECTION]: The skill defines a naming convention for plan files based on user-provided task descriptions. It instructs the agent to save these files in the project root. This is a standard project management function and follows typical file system usage patterns for development agents without introducing unnecessary privilege or access risks.

plan-writing