Add UAT guardrails to pipeline YAML and fix playbook test

- animation.yaml: enforce audio architecture + provider comparison in
  proposal stage, Layer 3 skill gate + clip duration in assets stage
- video-reference-analyst.md: Step 6 is now a hard redirect forcing
  stage-by-stage pipeline execution
- Fix test_compatible_with_manifest for nested compatible_playbooks dict

pipeline_defs/animation.yaml

@@ -95,10 +95,14 @@ stages:
       - Production plan reflects actual tool availability from preflight
       - Cost estimate is itemized and honest
       - Free/local tool options are highlighted where applicable
+      - "Audio architecture decided: single narrator / character dialogue / narrator+characters"
+      - "Provider choice presented as comparison table with costs — agent recommends but user decides"
     success_criteria:
       - Schema-valid proposal_packet with at least 3 concept_options
       - selected_concept includes animation_mode and reuse_strategy
+      - selected_concept includes audio_architecture (single_narrator | character_dialogue | narrator_plus_characters)
       - cost_estimate has itemized line_items with per-tool costs
+      - cost_estimate includes provider_comparison table with at least 2 provider options
       - approval.status is "approved" or "approved_with_changes" before proceeding
     sub_stages:
       - name: sample
@@ -191,9 +195,12 @@ stages:
       - Reusable motifs and templates are prepared and referenced
       - Missing tool paths are surfaced honestly
       - Total cost within approved budget from proposal_packet
+      - "Layer 3 skills read for EVERY generation tool before writing prompts (check agent_skills field)"
+      - "Clip duration maximized (prefer 10s over 5s) to reduce API calls and cost"
     success_criteria:
       - Schema-valid asset_manifest artifact
       - All referenced asset files exist on disk
+      - layer3_skills_read list is present and includes all tools used for generation
 
   - name: edit
     skill: pipelines/animation/edit-director

skills/meta/video-reference-analyst.md

@@ -327,13 +327,33 @@ adjust: voice, visual style, pacing, music, colors."
 Iterate on sample feedback until approved. Store samples at:
 `projects/<name>/assets/sample/sample_v{N}.mp4`
 
-### Step 6: Enter Pipeline
+### Step 6: Enter Pipeline (HARD REDIRECT)
 
-After sample approval, enter the appropriate pipeline with:
+After sample approval, the agent MUST enter the pipeline. This is not optional.
+
+**Mandatory steps:**
+1. Read the pipeline manifest: `pipeline_defs/animation.yaml` (or whichever
+   pipeline matches the production type)
+2. Execute **stage by stage** in order — research → proposal → script →
+   scene_plan → assets → edit → compose → publish
+3. Before EACH stage, read its director skill from
+   `skills/pipelines/<pipeline>/<stage>-director.md`
+4. Produce the required artifacts at each stage
+5. Hit every checkpoint where `checkpoint_required: true`
+6. Get user approval where `human_approval_default: true`
+
+**Do NOT collapse stages.** Do not jump from "user approved proposal" to
+"generate all assets." The pipeline stages exist to enforce quality gates,
+artifact dependencies, and review checkpoints. Skipping them is a governance
+violation.
+
+**Context to carry into the pipeline:**
 - VideoAnalysisBrief as grounding context in the research/proposal stage
 - User's chosen variant as the approved direction
 - Sample feedback incorporated into the brief
 - All creative differentiation decisions recorded in the decision_log
+- Audio architecture and voice casting decisions from Step 3
+- Layer 3 skills already read from Step 4b
 
 The pipeline takes over from here. The VideoAnalysisBrief travels alongside the
 standard artifacts, providing reference grounding at every stage.

tests/contracts/test_phase3_contracts.py

@@ -239,7 +239,10 @@ class TestStylePlaybooks:
     def test_compatible_with_manifest(self):
         manifest = load_pipeline("animated-explainer")
         available = list_playbooks()
-        for name in manifest.get("compatible_playbooks", []):
+        compat = manifest.get("compatible_playbooks", {})
+        # compatible_playbooks is a dict with recommended/also_works lists
+        playbook_names = compat.get("recommended", []) + compat.get("also_works", [])
+        for name in playbook_names:
             assert name in available, f"Manifest references unavailable playbook: {name}"