Add reference video input analysis workflow

skills/pipelines/cinematic/proposal-director.md

@@ -20,6 +20,32 @@ You are the **Proposal Director** for a cinematic video (trailers, brand films,
 
 ## Process
 
+### Step 0: Check for Reference Video Context
+
+Before starting proposal work, check if a VideoAnalysisBrief exists for this project.
+
+**When a VideoAnalysisBrief is present — Reference-Aware Cinematic Concept Design:**
+
+**HARD RULE: No carbon copies.** Each concept option MUST:
+1. Name at least ONE cinematic element it keeps from the reference (mood, pacing, color palette, shot language)
+2. Name at least ONE element it changes (emotional arc, visual treatment, subject matter, sound design)
+3. Explain WHY the change creates a different emotional impact
+
+**Cinematic differentiation patterns:**
+
+| Pattern | Example |
+|---------|---------|
+| **Same mood, different subject** | Reference: dark sci-fi mood → Ours: same darkness applied to deep ocean |
+| **Same subject, different emotional arc** | Reference: tension→reveal → Ours: wonder→scale |
+| **Same pacing, different visual language** | Reference: handheld raw → Ours: locked-off geometric |
+| **Same color world, different lighting** | Reference: warm golden hour → Ours: warm but tungsten/interior |
+
+**Mandatory Sample Protocol:** After concept approval, produce a 10-15 second cinematic
+sample BEFORE full production. This is critical for cinematic work — mood mismatches are
+expensive to fix downstream. Present with visual + audio + music.
+
+**When no VideoAnalysisBrief is present:** Skip this step and proceed normally.
+
 ### Step 1: Absorb the Research
 
 Read the `research_brief` thoroughly. Extract:

skills/pipelines/cinematic/research-director.md

@@ -18,6 +18,40 @@ Unlike explainer research (which focuses on facts, data, and content gaps), cine
 
 ## Process
 
+### Step 0: Check for Reference Video Context
+
+Before starting research, check if a VideoAnalysisBrief exists for this project. If it
+does, this is a reference-driven production — the user provided a video they want to
+riff on.
+
+**When a VideoAnalysisBrief is present:**
+
+1. Read it thoroughly. Extract:
+   - `content_analysis.topics` — research these topics for accuracy
+   - `content_analysis.key_claims` — verify these claims via web search
+   - `style_profile` — note the cinematic language (color palette, camera movements, lighting)
+   - `structure_analysis.scenes` — understand the shot language and emotional arc
+   - `replication_guidance.creative_differentiation_seeds` — these are your concept seeds
+   - `replication_guidance.key_elements_to_replicate` — preserve these in proposals
+
+2. Your research focus SHIFTS:
+   - Standard research: "What visual/emotional language fits this subject?"
+   - Reference-driven research: "What cinematic approach would DIFFERENTIATE us from the
+     reference while keeping the elements the user loved?" + "What mood/tone territory
+     is adjacent but unexplored?"
+
+3. In the research_brief, add a `reference_context` section:
+   - The reference's cinematic language (shot types, pacing, color palette)
+   - What emotional territory it occupies
+   - Adjacent emotional territories we could explore instead
+   - How the reference's visual approach could be evolved or reinterpreted
+
+4. The `angles_discovered` should explicitly position against the reference:
+   - "The reference uses X mood/palette/pacing. We could try Y which creates
+     [different emotional impact] because [research finding]."
+
+**When no VideoAnalysisBrief is present:** Skip this step and proceed normally.
+
 ### Step 1: Classify the Brief
 
 Before searching, extract from the user's request: