Short-Form Eval — Orchestrator

Feedback-loop skill — closes the brief → publish → score → pattern-log loop for a published short-form video. Scores hook + hold + save/share + comment-quality fidelity against the brief-shortform hypothesis. Gap-gate consumes its outputs to decide what the stack should learn next.

Core Question: "Did the published video survive contact with the platform — and what's the signal-bearing pattern this cycle adds to the log?"

Why, methodology, refutability, rubric-revision discipline, when NOT to use: references/playbook.md [PLAYBOOK].

Critical Gates

1. Provisional rubric, not locked. references/rubric.md is v0.1, provisional. Mandatory revision after cycle 2-3 against real variance. Encode per-cycle drift in the artifact, never silently in the rubric file.
2. Cycle 1 = 70% observation / 30% scoring. First cycle leans toward describing what you saw; later cycles harden scoring as variance accumulates.
3. Brief + reference catalog must exist. No platform-intel reference → BLOCKED. No brief-shortform hypothesis → BLOCKED. The eval scores a fidelity claim against known patterns; missing either side reduces the run to vibes.
4. No fabricated metrics. Every engagement number, completion rate, hold-curve point, save/share count, and sample-size claim cites the URL or panel screenshot it came from. Critic rubric #1 fails the artifact otherwise.
5. Pattern-log entries are atomic. One cycle = one pattern-log entry block (claim · evidence · refutability · expiry). Free-form prose patterns are unusable downstream.

Quality Gate