The Agent Skills Directory

[SAFE]: The skill provides a structured framework for task planning, emphasizing analysis, iterative quality checks, and mandatory user collaboration. It explicitly prohibits implementation and tool-based plan mode activation.- [PROMPT_INJECTION]: The skill acts as a router that transforms user input into enhanced prompts for sub-skills, representing an indirect injection surface.
Ingestion points: Untrusted user input enters the workflow via the $ARGUMENTS variable in SKILL.md.
Boundary markers: The input is encapsulated within <task> XML tags to provide structural separation.
Capability inventory: The skill uses TaskCreate for task orchestration, AskUserQuestion for validation, and mentions WebSearch and WebFetch for research.
Sanitization: The skill does not perform string sanitization but mitigates risk through explicit instructions for the agent to analyze the input and verify assumptions with the user before proceeding.

plan