The Agent Skills Directory

[SAFE]: The skill provides a robust framework for exploratory ideation and design, emphasizing user intent and constraint exploration.
[COMMAND_EXECUTION]: The skill uses logical operations such as READ and WRITE for interaction with the local filesystem, specifically to load project context and save design artifacts to docs/plans/designs/. These operations are well-scoped and consistent with the skill's documented purpose.
[INDIRECT_PROMPT_INJECTION]: The skill acts on user-provided ideas and existing project context, which constitutes an ingestion surface for untrusted data. However, the requirement to ask clarifying questions one at a time and the explicit guardrail preventing implementation without design approval serve as effective process-level boundaries.

oma-brainstorm