The Agent Skills Directory

[SAFE]: The skill instructions and associated files demonstrate clear, productivity-focused logic. There are no detections of obfuscation, credential harvesting, unauthorized network operations, or persistence mechanisms. The skill relies on standard tool interactions to perform its tasks.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied plans, designs, or ideas, which constitutes an attack surface for indirect instructions. However, the structured workflow involving a design tree and discrete rounds of questioning limits the impact of potential injections.
Ingestion points: The skill accepts a 'plan, decision, or idea' from the user for analysis.
Boundary markers: None explicitly defined in the instructions to separate user content from system logic.
Capability inventory: The skill is capable of looking up facts in the environment and invoking subagents for reading long context.
Sanitization: No explicit validation or filtering of the user-provided plan is mentioned.

grill-me