The Agent Skills Directory

[SAFE]: No malicious patterns were detected. The skill focuses on workflow management for multi-turn objectives.
[PROMPT_INJECTION]: The skill contains specific defensive instructions for the agent to treat "Thread goal" metadata as operational objectives rather than user-authored instructions designed to bypass safety policies. This is a positive security practice.
[INDIRECT_PROMPT_INJECTION]: While the skill establishes a surface for processing user-defined goals (the goal parameter in long_task), it provides clear boundary instructions and acceptance criteria to ensure the agent maintains state and does not interpret input as a protocol override.

long-goal