Activates the inline evaluation pipeline for Unknown Creatives Coach sessions. When active, every coaching response is evaluated against pre-written evals before the user sees it. Supports two modes: fork-based (primary, for interactive sessions) and named subagent fallback (for headless/SDK contexts). Use this skill when: testing the coach agent's quality, running a coaching session with quality assurance enabled, or when the plugin maintainer wants to evaluate coach performance. Triggers on: "evaluate session", "enable quality gate", "run with evaluation", "test the coach", "enable eval pipeline", or any request to activate the evaluation system during a coaching session.
How this skill is triggered — by the user, by Claude, or both
Slash command
/unknown-creatives-coach:evaluate-sessionThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Activates the evaluation pipeline for the current coaching session. When active, every coaching response passes through the judge loop before reaching the user.
Activates the evaluation pipeline for the current coaching session. When active, every coaching response passes through the judge loop before reaching the user.
When this skill is invoked:
unknown-creatives-coach-knowledge/coach-feedback.md if it exists — context on recent performance patterns.unknown-creatives-coach-knowledge/eval-health.md if it exists — flags any evals under review.Use when running in an interactive Claude Code session where forks are available.
For each coaching response:
agents/orchestrator.md for the full pipeline instructions."unknown-creatives-coach-knowledge/coach-feedback.md.Use when forks are unavailable (headless mode, Agent SDK, older Claude Code versions).
For each coaching response:
The coaching session (you, the consultant) never sees:
You receive only behavioral feedback: "Your response asked two questions — ask one." Never: "You failed the one-question-at-a-time eval."
The pipeline stays active for the entire session once activated. At session end, the orchestrator produces a session-end report if any defaults occurred.
To review pipeline health:
unknown-creatives-coach-knowledge/eval-health.md for flagged evals and threshold metricsunknown-creatives-coach-knowledge/coach-feedback.md for recurring patterns in judge feedbackProvides a checklist for code reviews covering functionality, security, performance, maintainability, tests, and quality. Use for pull requests, audits, team standards, and developer training.
npx claudepluginhub doterodesign/unknown-creatives-coach --plugin unknown-creatives-coach