dojo

Use when an Agent Skill needs behavioral evaluation, hardening, trigger tuning, isolation auditing, or evidence before release, or when the user explicitly asks to create a skill through Dojo. Runs coverage-scaled baseline and skilled scenarios, pressure-tests discipline skills, holds out graduation cases, checks routing collisions, and records proof under ~/.dojo. Responds to "dojo", "test this skill", "prove this skill works", "skill evals", "harden this SKILL.md", and "why is this skill triggering". Does not own ordinary first-draft skill creation unless Dojo is explicitly requested.

Extracts portable taste from any accessible medium into an evidence-traced profile and downstream production brief. Use when the user asks to extract, distill, reverse-engineer, or capture the style, taste, feel, visual language, voice, or character of an interface, image, text, video, music, object, brand, or experience so another agent can create new work with the same character without copying source content. Also use for cross-medium style translation. Refuse to fabricate a profile when the source cannot be inspected. Not for directly generating the final artifact unless the user asks for both extraction and creation.

Extracts portable taste from any accessible medium into an evidence-traced profile and downstream production brief. Use when the user asks to extract, distill, reverse-engineer, or capture the style, taste, feel, visual language, voice, or character of an interface, image, text, video, music, object, brand, or experience so another agent can create new work with the same character without copying source content. Also use for cross-medium style translation. Refuse to fabricate a profile when the source cannot be inspected. Not for directly generating the final artifact unless the user asks for both extraction and creation.