The Agent Skills Directory

[SAFE]: The skill defines a robust protocol for 'Ask compete' workflows. It emphasizes security best practices through candidate isolation, preventing cross-lane data leakage, and requiring local deterministic proof for all feature claims. All external tools mentioned (Brave Search, GitHub Search, Dogpile) are standard search services used for legitimate research during the competition process. The included sanity script and agentic evaluation fixture are benign validation tools that ensure the skill's markdown structure and dependencies adhere to the project's orchestration standards.

best-practices-competition