Slice 2 — Skill Consolidations Implementation Plan

For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (- [ ]) syntax for tracking.

Goal: Consolidate 5 sets of overlapping skills in videoclaw v3 to reduce duplication, sharpen the user-facing entry points, and keep the catalog honest about which skills are canonical vs alias.

Architecture: Skill-markdown surgery, not source code. Each consolidation is one self-contained commit: read the affected SKILL.md files, fold or stub or delete, update skills/catalog.json, run guardrails (check:skill-frontdoor + check:cleanroom-docs + npm test). The catalog already declares the alias structure (bunty, davendra-presenter, nex-presenter → aliasOf: brand-presenter); what's missing is the actual SKILL.md reduction.

Tech Stack: Markdown + JSON. No TS code changes. Existing guardrails:

• check:skill-frontdoor (bash, ignore-list driven)
• check:cleanroom-docs (bash, pattern-driven)
• check:movie-director-wrappers (bash, syntax-checks skills/movie-director/scripts/*.sh — unaffected)
• src/tests/skills-hygiene.test.ts (node:test — read assertions before editing skills)
• src/tests/package-scripts.test.ts (node:test — asserts skills/davendra-presenter/assets/davendra_intro_1.jpg + skills/nex-presenter/assets/nex_intro_1.jpg ship in the npm package; the asset directories must survive)

Source spec: docs/superpowers/specs/2026-05-25-videoclaw-v3-unification-design.md §4 Slice 2.

Effort target: ~3 days, 5 commits.

---

File Structure

Files to create:

• skills/bunty/brand-profile.json — per-presenter parameters (character_id, voice, brand assets, intro/outro)
• skills/davendra-presenter/brand-profile.json — same
• skills/nex-presenter/brand-profile.json — same

Files to modify:

• skills/brand-presenter/SKILL.md — make parametric on placeholders that the child skill resolves from its brand-profile.json
• skills/bunty/SKILL.md — reduce ~560 lines → ~25-line stub that points at brand-presenter + names its brand-profile
• skills/davendra-presenter/SKILL.md — reduce ~820 lines → ~25-line stub
• skills/nex-presenter/SKILL.md — reduce ~842 lines → ~25-line stub
• skills/video-framework/SKILL.md — augment with creative-brief's 7-question intake
• skills/movie-director/SKILL.md — fold director-video's unique content (mostly already covered)
• skills/seedance-prompts/SKILL.md — add music-video subsection
• skills/catalog.json — remove deleted entries, update alias declarations
• scripts/check-skill-frontdoor.sh — REMOVE the 3 presenter SKILL.mds from ignore_paths (after Commit 1 they no longer invoke Python directly)

Files to delete:

• skills/video-replicator/SKILL.md (DEMOTE — strip description so it doesn't auto-trigger; keep the directory + references/ + scripts/ which other skills depend on)
• skills/director-video/SKILL.md (after content folded into movie-director; delete the whole directory if nothing else references it)
• skills/creative-brief/SKILL.md (after content folded into video-framework; delete the whole directory)
• skills/seedance-music-video-prompts/SKILL.md (after merged into seedance-prompts; delete the whole directory)

Asset directories that MUST survive: skills/davendra-presenter/assets/, skills/nex-presenter/assets/ (asserted in package-scripts.test).

Guardrails to keep green after every commit:

• npm test (490+ tests)
• npm run check:skill-frontdoor
• npm run check:cleanroom-docs
• npm run check:release-readiness-lite at slice end

---

Commit Plan (5 commits)

Task 1: Presenter family — parametric brand-presenter + 3 stubs (Commit 1)

Goal: Eliminate ~2200 lines of duplicate workflow across bunty / davendra-presenter / nex-presenter SKILL.mds. Make brand-presenter the parametric base; reduce the children to thin profile stubs.

Files:

• Read: skills/brand-presenter/SKILL.md, skills/bunty/SKILL.md, skills/davendra-presenter/SKILL.md, skills/nex-presenter/SKILL.md, skills/catalog.json

• Create: skills/bunty/brand-profile.json, skills/davendra-presenter/brand-profile.json, skills/nex-presenter/brand-profile.json

• Modify: skills/brand-presenter/SKILL.md, skills/bunty/SKILL.md, skills/davendra-presenter/SKILL.md, skills/nex-presenter/SKILL.md, skills/catalog.json, scripts/check-skill-frontdoor.sh

• [ ] Step 1.1: Read all four SKILL.md files end-to-end

Run:

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
wc -l skills/brand-presenter/SKILL.md skills/bunty/SKILL.md skills/davendra-presenter/SKILL.md skills/nex-presenter/SKILL.md
cat skills/brand-presenter/SKILL.md

Then read each presenter SKILL.md in full. The 3 children duplicate substantial workflow; the goal is to identify their variation points (what's different between bunty/davendra/nex) vs the common workflow (what they share).

Expected variation points (extract these per presenter):

• presenter_name — "Bunty" / "Davendra" / "Nex"
• character_id — bunty=97, davendra=?, nex=?
• host_persona — short description (e.g., "3D Pixar-style executive presenter")
• intro_asset — path to the intro image
• outro_asset — path to the outro image
• voice_id — TTS voice
• subject_domains — what the presenter typically covers (cricket / corporate / tech)
• trigger_phrases — what user inputs route to this presenter (e.g., "cricket scorecard", "exec presenter", "tech commentator")

Everything else is workflow that lives in brand-presenter/SKILL.md.

• [ ] Step 1.2: Define the brand-profile.json schema

Create schemas/video/brand-profile.schema.json:

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "brand-profile",
  "description": "Per-brand parameters that specialize brand-presenter for a specific host (bunty/davendra/nex/...).",
  "type": "object",
  "required": ["presenterName", "characterId", "hostPersona", "triggerPhrases"],
  "properties": {
    "presenterName": { "type": "string", "description": "Human-facing host name (e.g., 'Bunty', 'Davendra', 'Nex')." },
    "characterId": { "type": ["integer", "string"], "description": "Go Bananas character ID for identity-locking generations." },
    "hostPersona": { "type": "string", "description": "Short descriptor of the host's visual + tonal style." },
    "subjectDomains": { "type": "array", "items": { "type": "string" }, "description": "Topics the presenter is typically used for." },
    "triggerPhrases": { "type": "array", "items": { "type": "string" }, "description": "Phrases in user input that should route to this brand profile." },
    "introAsset": { "type": "string", "description": "Relative path to the intro image inside this skill's assets/." },
    "outroAsset": { "type": "string", "description": "Relative path to the outro image inside this skill's assets/." },
    "voiceId": { "type": "string", "description": "TTS voice identifier (provider-specific)." },
    "voiceNotes": { "type": "string", "description": "Free-form notes about voice tone, pacing, accent." },
    "scriptStyle": { "type": "string", "description": "How the presenter's commentary script is shaped (e.g., 'cricket play-by-play', 'executive summary')." }
  },
  "additionalProperties": false
}

• [ ] Step 1.3: Write the three brand-profile.json files

For each presenter, extract the variation points identified in Step 1.1 into brand-profile.json.

Create skills/bunty/brand-profile.json (example shape — replace values with what bunty's SKILL.md actually documents):

{
  "presenterName": "Bunty",
  "characterId": 97,
  "hostPersona": "Cartoon Indian commentator in an orange blazer; warm, excitable, deeply knowledgeable about cricket.",
  "subjectDomains": ["cricket", "play-cricket scorecards", "match-day analysis"],
  "triggerPhrases": [
    "Match Day Analysis with Bunty",
    "cricket scorecard",
    "play-cricket URL",
    "Bunty"
  ],
  "introAsset": "assets/bunty_intro.jpg",
  "outroAsset": "assets/bunty_outro.jpg",
  "voiceId": "EXTRACT_FROM_SKILL_MD",
  "voiceNotes": "EXTRACT_FROM_SKILL_MD — tone, accent, pacing notes",
  "scriptStyle": "Cricket play-by-play with English/Hindi colloquialisms; references batting averages, partnerships, run rates."
}

For skills/davendra-presenter/brand-profile.json — extract from davendra-presenter's SKILL.md:

{
  "presenterName": "Davendra",
  "characterId": "EXTRACT_FROM_SKILL_MD",
  "hostPersona": "3D Pixar-style executive presenter; measured, professional, deck-style commentary.",
  "subjectDomains": ["corporate decks", "executive summaries", "investor updates"],
  "triggerPhrases": [
    "executive presenter",
    "exec deck",
    "Davendra",
    "corporate narrated video"
  ],
  "introAsset": "assets/davendra_intro_1.jpg",
  "outroAsset": "assets/davendra_outro_1.jpg",
  "voiceId": "EXTRACT_FROM_SKILL_MD",
  "voiceNotes": "EXTRACT_FROM_SKILL_MD",
  "scriptStyle": "Executive summary — concise, structured, slide-anchored narration."
}

For skills/nex-presenter/brand-profile.json:

{
  "presenterName": "Nex",
  "characterId": "EXTRACT_FROM_SKILL_MD",
  "hostPersona": "3D Pixar-style tech commentator; technical depth with accessible delivery.",
  "subjectDomains": ["tech reviews", "product walkthroughs", "developer commentary"],
  "triggerPhrases": [
    "tech commentator",
    "tech review",
    "Nex",
    "product walkthrough video"
  ],
  "introAsset": "assets/nex_intro_1.jpg",
  "outroAsset": "assets/nex_outro_1.jpg",
  "voiceId": "EXTRACT_FROM_SKILL_MD",
  "voiceNotes": "EXTRACT_FROM_SKILL_MD",
  "scriptStyle": "Tech commentary — explanatory, concrete, with concrete code/spec references."
}

EXTRACT_FROM_SKILL_MD placeholders mean: open the corresponding SKILL.md and find the actual value. Don't ship literal EXTRACT_FROM_SKILL_MD strings; they're TODO markers.

• [ ] Step 1.4: Update brand-presenter/SKILL.md to be parametric

The current skills/brand-presenter/SKILL.md is 85 lines and likely already generic. Modify it to reference brand-profile.json explicitly. Add this section near the top (after the YAML frontmatter):

## Brand profile

This skill is parametric. Each child skill (`bunty`, `davendra-presenter`,
`nex-presenter`, future presenters) provides a `brand-profile.json` that
specializes this workflow for a specific host.

Schema: [`schemas/video/brand-profile.schema.json`](../../schemas/video/brand-profile.schema.json).

When this skill is invoked via a child alias, the orchestrator should:
1. Load the child's `brand-profile.json`.
2. Substitute `{{presenterName}}`, `{{characterId}}`, `{{hostPersona}}`,
   `{{introAsset}}`, `{{outroAsset}}`, etc. throughout the workflow below.
3. Use `triggerPhrases` to confirm intent routing was correct.

Keep the rest of the workflow as-is.

• [ ] Step 1.5: Reduce bunty/SKILL.md to a stub

Replace the contents of skills/bunty/SKILL.md with this stub (about 25 lines):

---
name: bunty
description: |
  "Match Day Analysis with Bunty" — cricket scorecard or play-cricket URL
  to a narrated video with Bunty (cartoon Indian commentator in orange
  blazer, character_id=97). Triggers on: cricket scorecards, play-cricket
  URLs, "Match Day Analysis", "Bunty".
aliasOf: brand-presenter
---

# Bunty (brand-presenter specialization)

Bunty is a `brand-presenter` specialization. The full workflow lives in
[`skills/brand-presenter/SKILL.md`](../brand-presenter/SKILL.md). This
stub provides the brand parameters that specialize the parent workflow.

**Brand profile:** [`brand-profile.json`](./brand-profile.json)

**Trigger:** When a user provides a cricket scorecard PDF, a play-cricket
URL, or asks for "Match Day Analysis", route to `brand-presenter` with
this profile loaded.

**Assets:** `assets/` holds Bunty's intro/outro reference images +
auxiliary character-sheet materials. The brand-profile.json points at
them.

**Why this stub exists:** prior to Slice 2 (2026-05-25 cutover), Bunty's
full workflow lived in this SKILL.md as a 560-line copy of the
brand-presenter pipeline. The duplication is now consolidated in
`brand-presenter` with this profile-driven specialization.

• [ ] Step 1.6: Reduce davendra-presenter/SKILL.md to a stub

Replace the contents of skills/davendra-presenter/SKILL.md with:

---
name: davendra-presenter
description: |
  Create a professional narrated video from a PDF slide deck featuring
  **Davendra**, a 3D Pixar-style executive presenter. Triggers on:
  "executive presenter", "exec deck video", "Davendra", "corporate
  narrated video".
aliasOf: brand-presenter
---

# Davendra Presenter (brand-presenter specialization)

Davendra is a `brand-presenter` specialization. The full workflow lives
in [`skills/brand-presenter/SKILL.md`](../brand-presenter/SKILL.md).
This stub provides the brand parameters.

**Brand profile:** [`brand-profile.json`](./brand-profile.json)

**Trigger:** When a user wants a narrated corporate / executive video
with a 3D Pixar-style host, route to `brand-presenter` with this profile
loaded.

**Assets:** `assets/davendra_intro_1.jpg` and the corresponding outro
image. The brand-profile.json points at them.

**Why this stub exists:** prior to Slice 2, the full ~820-line workflow
duplicated brand-presenter. Now consolidated via `brand-profile.json`.

• [ ] Step 1.7: Reduce nex-presenter/SKILL.md to a stub

Same pattern. Replace skills/nex-presenter/SKILL.md with:

---
name: nex-presenter
description: |
  Create a professional narrated video from a PDF slide deck featuring
  **Nex**, a 3D Pixar-style tech commentator. Triggers on: "tech
  commentator", "tech review video", "Nex", "product walkthrough".
aliasOf: brand-presenter
---

# Nex Presenter (brand-presenter specialization)

Nex is a `brand-presenter` specialization. The full workflow lives in
[`skills/brand-presenter/SKILL.md`](../brand-presenter/SKILL.md). This
stub provides the brand parameters.

**Brand profile:** [`brand-profile.json`](./brand-profile.json)

**Trigger:** When a user wants a narrated tech / product video with a
3D Pixar-style host, route to `brand-presenter` with this profile loaded.

**Assets:** `assets/nex_intro_1.jpg` and the corresponding outro image.
The brand-profile.json points at them.

**Why this stub exists:** prior to Slice 2, the full ~842-line workflow
duplicated brand-presenter. Now consolidated via `brand-profile.json`.

• [ ] Step 1.8: Verify catalog.json already declares the alias structure

Run:

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
jq '.skills[] | select(.id == "bunty" or .id == "davendra-presenter" or .id == "nex-presenter" or .id == "brand-presenter")' skills/catalog.json

Expected: brand-presenter has role: canonical-entry and specializations: [davendra-presenter, nex-presenter, bunty]. The three children have status: alias, aliasOf: brand-presenter. If anything's missing or wrong, fix it.

• [ ] Step 1.9: Remove the 3 presenter SKILL.mds from check-skill-frontdoor ignore list

The ignore list at scripts/check-skill-frontdoor.sh lines 16-22 still lists bunty, davendra-presenter, nex-presenter as ignored because they invoked Python pipeline scripts directly. After this commit, the stubs no longer reference those scripts (the brand-presenter parent handles invocation). Remove them.

Edit scripts/check-skill-frontdoor.sh:

# Before:
ignore_paths=(
  "skills/seedance-prompts/SKILL.md"
  # Presenter skills legitimately invoke their bundled Python pipeline
  # under skills/video-replicator/scripts/ (ported in Phase 4a from the
  # canonical workspace). Those scripts ARE the implementation; calling
  # them is the right way. The scanner predates the pipeline import.
  "skills/bunty/SKILL.md"
  "skills/davendra-presenter/SKILL.md"
  "skills/nex-presenter/SKILL.md"
  # movie-director references the legacy script paths in its
  # references/ subdir; those docs are historical-reference-style.
  # The SKILL.md itself uses the vclaw CLI surface.
)

# After:
ignore_paths=(
  "skills/seedance-prompts/SKILL.md"
  # movie-director references the legacy script paths in its
  # references/ subdir; those docs are historical-reference-style.
  # The SKILL.md itself uses the vclaw CLI surface.
)

• [ ] Step 1.10: Run guardrails

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
npm run build && npm test 2>&1 | tail -5
npm run check:skill-frontdoor
npm run check:cleanroom-docs

Expected: tests green (493 pass), skill-frontdoor passes (the stubs no longer reference Python directly), cleanroom-docs passes. If a test fails, the assertion is likely against old SKILL.md content — check and adapt.

• [ ] Step 1.11: Commit

git add skills/brand-presenter/ skills/bunty/ skills/davendra-presenter/ skills/nex-presenter/ schemas/video/brand-profile.schema.json scripts/check-skill-frontdoor.sh skills/catalog.json
git commit -m "Slice 2: presenter family parametric — brand-profile.json + 3 stubs (~2200 lines deleted)"

---

Task 2: video-replicator demotion (Commit 2)

Goal: video-framework becomes the sole video front door; video-replicator stops auto-firing on user intent but stays as a reference document for the 7-mode legacy surface.

Files:

• Read: skills/video-replicator/SKILL.md (371 lines), skills/video-framework/SKILL.md, skills/catalog.json

• Modify: skills/video-replicator/SKILL.md (strip description), skills/catalog.json

• [ ] Step 2.1: Read both SKILL.mds

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
cat skills/video-framework/SKILL.md
head -60 skills/video-replicator/SKILL.md

Confirm: video-replicator currently has a YAML frontmatter description: field that auto-triggers on "make me a video" type intents — same intents video-framework should handle.

• [ ] Step 2.2: Strip video-replicator's description (demote to reference-only)

Edit skills/video-replicator/SKILL.md. Modify the YAML frontmatter:

---
name: video-replicator
description: |
  Reference-only documentation of the legacy 7-mode video-replicator
  surface. Do NOT use this skill to trigger work — use
  `video-framework` instead. This document is preserved as historical
  context for the modes and their parameters; the new user-facing entry
  point is `video-framework`.
---

The key change: the description explicitly says "Do NOT use this skill to trigger work" so Claude's auto-trigger matcher routes elsewhere. Add a banner at the top of the body content:

> **DEPRECATED FOR USER-FACING USE.** This skill's auto-trigger has been
> disabled in Slice 2. The user-facing video front door is now
> [`video-framework`](../video-framework/SKILL.md). This file remains as
> reference documentation for the legacy 7-mode surface.

Leave the rest of the file content intact for reference.

• [ ] Step 2.3: Update catalog.json

Change video-replicator's catalog entry from its current category to:

{
  "id": "video-replicator",
  "category": "video",
  "status": "deprecated-reference",
  "role": "reference",
  "supersededBy": "video-framework"
}

• [ ] Step 2.4: Run guardrails

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
npm run build && npm test 2>&1 | tail -5
npm run check:skill-frontdoor
npm run check:cleanroom-docs

Expected: green.

• [ ] Step 2.5: Commit

git add skills/video-replicator/SKILL.md skills/catalog.json
git commit -m "Slice 2: demote video-replicator to reference; video-framework is sole front door"

---

Task 3: Merge director-video into movie-director (Commit 3)

Goal: Eliminate the director-video skill; movie-director is the richer of the two and absorbs anything director-video had that movie-director doesn't.

Files:

• Read: skills/movie-director/SKILL.md, skills/director-video/SKILL.md, skills/catalog.json

• Modify: skills/movie-director/SKILL.md (fold in additive content if any), skills/catalog.json

• Delete: skills/director-video/ (entire directory after fold)

• [ ] Step 3.1: Diff the two SKILL.mds

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
diff -u skills/director-video/SKILL.md skills/movie-director/SKILL.md | head -100

Identify content in director-video/SKILL.md that's NOT in movie-director/SKILL.md. Examples to look for:

• Triggers/keywords movie-director doesn't catch

• Workflow steps movie-director doesn't document

• References to scripts/assets only in director-video/

• [ ] Step 3.2: Fold additive content into movie-director

If anything's additive, add it to skills/movie-director/SKILL.md in the appropriate section. If director-video is a strict subset of movie-director (likely), no edits to movie-director needed — proceed to deletion.

If you ADD content, mark it with a brief comment line so future maintainers see the lineage:

<!-- Folded in from skills/director-video/SKILL.md in Slice 2 (2026-05-25). -->

• [ ] Step 3.3: Check if anything else references director-video

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
grep -rln "director-video" skills/ src/ scripts/ docs/ 2>/dev/null | grep -v "/director-video/" | head -10

If anything references it (e.g., other SKILL.mds linking to it), update those references to point to movie-director instead.

• [ ] Step 3.4: Delete the director-video directory

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
rm -rf skills/director-video/

• [ ] Step 3.5: Update catalog.json

Remove the director-video entry entirely from skills/catalog.json. Also update movie-director's specializes field — it currently says specializes: director-video which becomes stale.

{
  "id": "movie-director",
  "category": "video",
  "status": "imported",
  "specializes": "video-framework",
  "role": "specialist"
}

(Or whatever specialization makes sense — pick video-framework as the parent since director-video is going away.)

• [ ] Step 3.6: Run guardrails

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
npm run build && npm test 2>&1 | tail -5
npm run check:movie-director-wrappers
npm run check:skill-frontdoor
npm run check:cleanroom-docs

Expected: green. check:movie-director-wrappers syntax-checks files under skills/movie-director/scripts/; unaffected by director-video deletion.

• [ ] Step 3.7: Commit

git add skills/movie-director/SKILL.md skills/catalog.json
git rm -r skills/director-video/
git commit -m "Slice 2: merge director-video into movie-director (deleted)"

---

Task 4: Fold creative-brief into video-framework (Commit 4)

Goal: creative-brief's 7-question intake becomes video-framework's intake mode. creative-brief skill goes away.

Files:

• Read: skills/creative-brief/SKILL.md, skills/video-framework/SKILL.md, skills/catalog.json

• Modify: skills/video-framework/SKILL.md (add intake section), skills/catalog.json

• Delete: skills/creative-brief/ (entire directory)

• [ ] Step 4.1: Read both files

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
cat skills/creative-brief/SKILL.md
cat skills/video-framework/SKILL.md

Identify creative-brief's 7-question intake structure. It's likely a numbered list of prompts the operator asks the user.

• [ ] Step 4.2: Add an "Intake mode" section to video-framework/SKILL.md

Append to skills/video-framework/SKILL.md (or insert in the appropriate location — read the file to decide):

## Intake mode (formerly `creative-brief`)

When the user's request is too vague to route directly, drop into intake
mode and ask the 7 questions below. Skip questions whose answer is
already clear from the request.

<!-- Folded in from skills/creative-brief/SKILL.md in Slice 2 (2026-05-25). -->

1. [QUESTION 1 FROM creative-brief/SKILL.md]
2. [QUESTION 2 ...]
3. [QUESTION 3 ...]
4. [QUESTION 4 ...]
5. [QUESTION 5 ...]
6. [QUESTION 6 ...]
7. [QUESTION 7 ...]

After answers are collected, map them to `vclaw video brief --project
... --title ... --intent ... --aspect-ratio ... --quality ...` flags
and proceed with the regular video-framework dispatch.

Extract the actual 7 questions from creative-brief/SKILL.md. Don't ship literal placeholders.

• [ ] Step 4.3: Check for references to creative-brief elsewhere

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
grep -rln "creative-brief" skills/ src/ scripts/ docs/ 2>/dev/null | grep -v "/creative-brief/" | head

Update any references to point at video-framework's intake mode.

• [ ] Step 4.4: Delete creative-brief directory

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
rm -rf skills/creative-brief/

• [ ] Step 4.5: Update catalog.json

Remove the creative-brief entry from skills/catalog.json.

• [ ] Step 4.6: Run guardrails

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
npm run build && npm test 2>&1 | tail -5
npm run check:skill-frontdoor
npm run check:cleanroom-docs

Expected: green.

• [ ] Step 4.7: Commit

git add skills/video-framework/SKILL.md skills/catalog.json
git rm -r skills/creative-brief/
git commit -m "Slice 2: fold creative-brief into video-framework intake mode (deleted)"

---

Task 5: Merge seedance-prompts variants (Commit 5)

Goal: seedance-music-video-prompts is a strict subset of seedance-prompts. Fold its content into the parent and delete the variant.

Files:

• Read: skills/seedance-prompts/SKILL.md, skills/seedance-music-video-prompts/SKILL.md, skills/catalog.json

• Modify: skills/seedance-prompts/SKILL.md (add music-video subsection), skills/catalog.json

• Delete: skills/seedance-music-video-prompts/ (entire directory)

• [ ] Step 5.1: Read both files

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
cat skills/seedance-prompts/SKILL.md
cat skills/seedance-music-video-prompts/SKILL.md

Identify what music-video-prompts uniquely covers (genres, beat-sync, lyric-overlay considerations, etc.).

• [ ] Step 5.2: Add a music-video subsection to seedance-prompts/SKILL.md

Append (or insert at the appropriate section break) to skills/seedance-prompts/SKILL.md:

## Music-video prompts (formerly seedance-music-video-prompts)

When the user wants a music video specifically — beat-synced visuals,
lyric-driven scene changes, performance shots — use the music-video
prompt patterns below in addition to the general Seedance prompt
patterns above.

<!-- Folded in from skills/seedance-music-video-prompts/SKILL.md in Slice 2 (2026-05-25). -->

[ACTUAL CONTENT EXTRACTED FROM seedance-music-video-prompts/SKILL.md]

Replace [ACTUAL CONTENT EXTRACTED ...] with the actual content. Don't ship the placeholder.

• [ ] Step 5.3: Delete seedance-music-video-prompts directory

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
rm -rf skills/seedance-music-video-prompts/

• [ ] Step 5.4: Update catalog.json

Remove the seedance-music-video-prompts entry (if present).

• [ ] Step 5.5: Run guardrails

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
npm run build && npm test 2>&1 | tail -5
npm run check:skill-frontdoor
npm run check:cleanroom-docs

Expected: green. Note: seedance-prompts/SKILL.md is in the cleanroom-docs ignore list — adding music-video content there won't trigger the guardrail.

• [ ] Step 5.6: Commit

git add skills/seedance-prompts/SKILL.md skills/catalog.json
git rm -r skills/seedance-music-video-prompts/
git commit -m "Slice 2: merge seedance-music-video-prompts into seedance-prompts (deleted)"

---

Final gate: full release-readiness-lite

After all 5 commits land, run the slice-completion gate:

cd /Users/davendrapatel/Documents/GitHub/videoclaw-v3
npm run check:release-readiness-lite 2>&1 | tail -10

Expected: release-readiness-lite checks passed.

If any guardrail fails:

• check:skill-frontdoor — likely a presenter stub still references Python; re-check the stubs and either reword to avoid the patterns or add the file back to the ignore list (with comment justifying it).
• check:cleanroom-docs — unlikely; the ignore list already covers docs/superpowers/ from Slice 1.
• npm test — likely a skills-hygiene test asserting a specific string in a SKILL.md you deleted/changed. Read the test, update the assertion to match v3 reality.

---

Failure modes + rollback paths

• Each commit is a single git revert. If any of the 5 commits causes issues downstream, git revert <sha> restores cleanly.
• Asset directories are sacred. Tests assert skills/davendra-presenter/assets/davendra_intro_1.jpg + skills/nex-presenter/assets/nex_intro_1.jpg ship in the npm package. NEVER delete those directories — only reduce the SKILL.md files inside.
• video-replicator's directory MUST survive. Other skills (the presenters) reference skills/video-replicator/scripts/... Python pipeline. Task 2 strips its SKILL.md description but leaves the scripts intact.
• catalog.json is hand-edited JSON. A trailing comma or missing field breaks parsing. After every edit, run jq . skills/catalog.json > /dev/null to validate before committing.
• Skill stubs use YAML frontmatter. The --- delimiters and name: / description: / aliasOf: fields must be exact for Claude's skill loader to pick them up. Test by running vclaw schema --json | jq '.commands | length' after each commit — number shouldn't drop (commands don't change in Slice 2; only skills do).

---

What ships after Slice 2

After all 5 commits:

• 3 presenter SKILL.mds reduced from ~560/820/842 → ~25 lines each (~2150 lines deleted)
• video-replicator no longer auto-triggers on user intent
• director-video deleted
• creative-brief deleted (intake folded into video-framework)
• seedance-music-video-prompts deleted (folded into seedance-prompts)
• catalog.json cleaner: fewer canonical entries, alias structure honest
• check-skill-frontdoor ignore list shorter

Skills count change: 53 → ~49 (drops director-video, creative-brief, seedance-music-video-prompts; presenters stay as stubs).

What does NOT ship:

• Python pipeline fold (Slice 3) — skills/video-replicator/scripts/ Python is still there
• Bun standalone surface collapse (Slice 4)
• MCP server + external skills pack (Slice 5)
• Migration of remaining ~40 vclaw.ts handlers to VclawError (incremental v3.x alpha work)

Next slice to plan: Slice 3 — Python fold (F2a) (3-4 weeks, biggest engineering piece).