@lalalic/markcut 2.2.7 → 2.2.9

package/skills/markcut/docs/templates/short-film/prompts/screenplay.md

@@ -1,101 +0,0 @@
-# Prompt: Screenplay — Three-Act Structure
-
-> Fill every `{placeholder}`, then execute in your own context.
-> Produces a scene-by-scene narrative plan with character arc, shot suggestions, and VO lines.
-
----
-
-You are writing a {duration_min}-minute {film_type} short film in a {tone} tone.
-
-Concept: **{concept}**
-
-Source material: {source_material}
-Language: {language}
-Voice(s): {voices}
-
-## Requirements
-
-### Three-act structure
-
-| Act | Function | Duration | Scenes |
-|---|---|---|---|
-| Teaser | Mood establisher, no context | 30–90s | 1 |
-| Act 1 — Setup | Character, world, inciting incident | 90s–2min | 3 |
-| Act 2 — Confrontation | Rising action, midpoint, darkest moment | 3–5min | 4–5 |
-| Act 3 — Resolution | Climax, falling action, resolution | 90s–2min | 3 |
-| Epilogue | Emotional close | 30–60s | 1 |
-
-### Protagonist rules
-- The protagonist must have a **desire** that drives the plot.
-- The **inciting incident** must happen by the end of Act 1.
-- The **midpoint** in Act 2 must shift the protagonist from reacting to acting.
-- The **darkest moment** must occur just before the climax.
-- The **climax** is where the protagonist confronts the central conflict — emotionally, not necessarily physically.
-- The **resolution** shows the new ordinary — how the protagonist has changed.
-
-### Scene format
-For each scene, provide:
-
-```
-### Act X.Y — Scene Title (INT/EXT. LOCATION - TIME)
-- Shot type: <wide/medium/close-up/POV/etc>
-- Camera movement: <static/push-in/handheld/etc>
-- Lighting: <golden hour/moonlight/neon/etc>
-- Mood: <one word>
-- Color palette: <cool/warm/desaturated/etc>
-- Visual prompt: <one-sentence TTI prompt with shot, subject, lighting, mood>
-- VO/Dialogue: <narration line or character dialogue>
-- Duration: <seconds>
-- Sound: <score/ambient/silence>
-```
-
-### VO rules
-- VO is minimal. The story is told through images.
-- One VO line per scene (max two). Each 10–30 words.
-- For dialogue, mark speaker: "JOHN: I never thought..."
-- No exposition dumps. Show, don't tell.
-
-## Output format
-
-```
-Title: <title>
-
-Logline: <one sentence>
-
-### Synopsis (3 sentences)
-1. <Beginning>
-2. <Middle>
-3. <End>
-
-### Teaser
-- Scene: <description>
-- Visual: <prompt>
-- VO: <line>
-- Duration: <s>
-
-### Act 1 — <title>
-
-#### Scene 1.1 — <name> (<location>)
-- Shot: <type>
-- Camera: <movement>
-- Lighting: <description>
-- Color: <palette>
-- Visual: <prompt>
-- VO: <line>
-- Duration: <s>
-- Sound: <note>
-
-#### Scene 1.2 ...
-...
-
-### Act 2 ...
-
-### Act 3 ...
-
-### Epilogue
-- Visual: <prompt>
-- VO: <line>
-- Duration: <s>
-
-Total estimated duration: <sum>
-```

package/skills/markcut/docs/templates/short-film/prompts/storyboard.md

@@ -1,54 +0,0 @@
-# Prompt: Scene → Markcut Shot
-
-> Fill every `{placeholder}`, then execute in your own context.
-> Run once per scene from the screenplay. Produces the exact markcut markdown for ONE scene.
-
----
-
-You are turning one scene of a {tone} short film into markcut markdown.
-
-Film: {film_title}
-Scene from screenplay:
-
-{screenplay_scene}
-
-## Rules
-
-### Visual prompt
-Write a detailed TTI prompt with ALL of these elements:
-1. **Shot type**: wide, medium, close-up, POV, over-the-shoulder, Dutch angle, establishing
-2. **Subject**: who or what is in the frame, what are they doing, what emotion
-3. **Composition**: where is the subject in the frame, foreground/background
-4. **Lighting**: specific light source and quality
-5. **Camera movement**: static, push-in, handheld, tracking, pan, tilt
-6. **Color palette**: warm, cool, desaturated, high contrast, etc.
-7. **Mood**: one word or short phrase
-8. **End with**: ", cinematic, film still, 4K, highly detailed"
-
-### Narration / dialogue
-- One `- script "..."` node containing the VO or dialogue line.
-- For dialogue: `"CHARACTER: line"`
-- Match the exact text from the screenplay.
-
-### Duration
-Match the screenplay's specified duration ±2s. Adjust if the VO line is longer/shorter.
-
-### Effect
-- `effect:fadeIn` for soft opens
-- `effect:zoomIn` for dramatic reveals or tension building
-- `effect:zoomOut` for reveals of context
-- No effect for static, contemplative scenes
-
-### Sound note (optional)
-Add a comment in the script or a separate note about sound design for this scene.
-
-## Output format for ONE scene
-
-```
-### <Act>.<Scene> — <Title> (<Location>)
-layout:parallel
-- image src:auto
-  prompt:"<SHOT TYPE> of <SUBJECT>, <COMPOSITION>, <LIGHTING>, <CAMERA MOVEMENT>, <COLOR>, <MOOD>, cinematic, film still, 4K"
-  duration:<s> effect:<effect>
-- script "<VO or dialogue>" duration:<s>
-```

package/skills/markcut/docs/templates/short-video/TEMPLATE.md

@@ -1,301 +0,0 @@
----
-name: short-video
-description: Turn a topic, text, image, or video into a punchy short-form vertical video (TikTok/Reels/Shorts) with animated captions, fast cuts, and BGM.
-when-to-use: explainers, listicles, quote videos, image showcases, highlight reels, faceless channels — 15–60s vertical videos
-engine: "@lalalic/markcut — run via `npx @lalalic/markcut`"
-aspect: "9:16 (1080×1920)"
----
-
-# Short Video Template
-
-Follow this file top to bottom. Read the markcut skill (`SKILL.md` → `docs/markdown-descriptive.md`) first if you have not.
-
-```mermaid
-flowchart LR
-    Input[📥 topic / text / image / video] --> Script[✍️ script.md<br/>hook + body + CTA]
-    Script --> Board[📋 scene breakdown]
-    Board --> Storyboard[📄 short.md]
-    BGM[🎵 bgm select] --> Storyboard
-    Media[🖼️ src:auto / source media] --> Storyboard
-    Storyboard --> Render[🎬 npx markcut render]
-    Render --> MP4[📺 final.mp4]
-    MP4 --> Reviewer[🔍 reviewer agent]
-    Reviewer -->|FAIL| Fix[🔧 fix]
-    Fix --> Storyboard
-    Reviewer -->|PASS| Done[✅ Done]
-```
-
-| Path | Runs in | Purpose |
-|---|---|---|
-| `TEMPLATE.md` | your context | everything |
-| `prompts/*.md` | your context | fill-in prompts you execute |
-| `agents/*.md` | separate session | subagent definitions |
-
-## 0. Prerequisites
-
-- `npx @lalalic/markcut` runnable
-- `ffmpeg`/`ffprobe` on PATH
-- ITT/TTV CLI (for `src:auto` background generation) — optional but recommended
-- For reviewer: image-understanding capability, STT CLI
-
-## 1. Inputs — collect before starting
-
-| Input | Required | Default | Notes |
-|---|---|---|---|
-| Topic / source material | **yes** | — | a topic to explain, a quote, an article, a story, or a URL to summarize |
-| Source media | no | — | optional: image file path, video file path, or folder of media to turn into a short |
-| Format | no | `explainer` | `explainer` (topic→text), `quote` (quote+background), `showcase` (image→narrated), `highlight` (video→best moments), `storytime` (short story) |
-| Language | no | en | narration language |
-| Style | no | `cinematic` | `cinematic` (dramatic), `energetic` (fast cuts, bright), `minimal` (clean, slow), `trendy` (meme-aware, casual) |
-| Target duration | no | 30s | 15–60s |
-| Voice | no |  | mlx-audio voice |
-| BGM mood | no | `upbeat` | mood for background music. BGM is **mandatory**. |
-
-**Rule:** For explainer/quote/storytime formats, generate all visuals via `src:auto prompt:"..."` — describe vivid, cinematic scenes. For showcase/highlight formats, use the provided source media and let the narration describe it.
-
-## 2. Scene grammar — mandatory structure
-
-### Overview
-
-```
-# video                          ← root: width:1080 height:1920 fps:30 layout:series
-│                                  subtitle:{fontSize:"56px",type:"Typewriter",fontFamily:"Arial Black"}
-│                                  
-│                                  transition:fade(0.3)
-├── - audio isBackground:true foreground:true src:bgm.mp3 volume:0.1
-│
-├── ## Hook                      ← 1–3s, FIRST FRAME HOOKS THE VIEWER
-│   layout:parallel
-│   - image/video src:auto prompt:"<arresting visual>" effect:zoomIn
-│   - subtitle src:"<hook line — one punchy sentence>" duration:3 type:Bounce
-│   - script "<hook narration>" (optional)
-│
-├── ## Body1                     ← each body scene = one key point
-│   layout:parallel
-│   - image/video src:auto prompt:"<scene visual>" effect:slideInRight
-│   - subtitle src:"<key point — short, scannable>" duration:5 type:Typewriter
-│   - script "<narration for this point>"
-│
-├── ## Body2                     ← 3–7 body scenes total
-│   ...                          ← each 3–8 seconds, fast-paced
-│
-└── ## CTA                       ← 3–5s, call to action
-    layout:parallel
-    - image/video src:auto prompt:"<closing visual>" effect:zoomIn
-    - subtitle src:"<CTA — like/subscribe/follow>" duration:4 type:Bounce
-    - script "<final words>"
-```
-
-### The Hook rule — first 2 seconds decide everything
-
-The hook is **the most important part of a short video**. If it doesn't grab attention in the first 2 seconds, the viewer scrolls past. Rules:
-
-- **Hook line**: one short, punchy sentence that creates curiosity, surprise, or emotional connection. Use questions ("Ever wondered why..."), bold claims ("Most people get this wrong"), or intriguing statements ("I tried X for 30 days.").
-- **Hook visual**: arresting, high-impact image/video that matches the hook line. Use `effect:zoomIn` for dramatic entrance.
-- **Hook duration**: 1–3 seconds max. Get to the content fast.
-- **No slow fade-ins, no logo animations, no intro music swell** — straight into the hook.
-
-### Body scenes — rapid, scannable
-
-- **3–7 body scenes**, each 3–8 seconds. Total body duration = target − hook − CTA.
-- **One key point per scene**. A single idea, fact, step, or argument.
-- **Text overlay (`subtitle`) is the star**: short, scannable, large font (48–64px). Use `type:Typewriter` or `type:Fade` for animated entrance.
-- **Visual background**: full-screen image/video that's relevant, cinematic, or metaphorical. Use `effect:slideInRight` or `effect:zoomIn` for scene transitions.
-- **Narration (`script`)**: optional but recommended. If used, keep it tight (5–15 words per scene). The subtitle text and narration should match closely but the subtitle is more scannable (shorter).
-
-### CTA scene
-
-- 3–5 seconds.
-- Clear call to action: "Follow for more", "Like if you agree", "Share this with someone who needs to hear it".
-- Visual: inviting, positive, or memorable.
-- Subtitle: `type:Bounce` for energetic CTA.
-
-### Scene timing template
-
-| Scene | Duration | Subtitle type | Effect | Visual |
-|---|---|---|---|---|
-| Hook | 1–3s | `Bounce` | `zoomIn` | arresting, high-impact |
-| Body×N | 3–8s each | `Typewriter` or `Fade` | `slideInRight` | scene-relevant, cinematic |
-| CTA | 3–5s | `Bounce` | `zoomIn` | closing, positive |
-
-## 3. Authoring rules — the short-form bar
-
-### Script writing
-
-- **Hook script**: 5–15 words. One sentence. Creates curiosity. No filler.
-- **Body scripts**: 5–15 words per scene. One idea per scene. Conversational, punchy.
-- **Total words**: target duration × 3 words/second (shorter-than-speech pace). A 30s video ≈ 90 words total.
-- **No introductions**: no "today I'm going to tell you about..." Get into the content immediately.
-- **Voice**: match the style. Cinematic = authoritative, confident. Energetic = fast, excited. Minimal = calm, slow. Trendy = casual, relatable.
-
-### Subtitle / caption text
-
-- **Subtitle is separate from script**: subtitle text is what appears on screen (scannable, shorter). Script is what TTS speaks (can be slightly longer). They should convey the same idea.
-- **Font size**: 48–64px. Short videos are watched on mobile — text must be large and legible.
-- **Max 10 words per subtitle frame**. If a script has more words, split into multiple subtitle scenes or keep the subtitle terse and let the script elaborate.
-- **Animated caption types**: `Typewriter` for reveal (most popular), `Bounce` for emphasis (hook, CTA), `Fade` for transitions, `Colorful` for energetic/fun content.
-- **Position**: subtitles auto-center by default. For safety (not covering faces), use `style:bottom:20%` on the subtitle node to push captions to the lower third.
-
-### Visuals (images/video backgrounds)
-
-- **Explainers & quotes**: use `src:auto prompt:"<vivid scene description>"` to generate backgrounds. Describe the mood, lighting, and composition (e.g., "cinematic shot of a library at twilight, warm lighting, shallow depth of field").
-- **Showcase**: use the source image as the background. If it's not full-screen, add a dark gradient overlay or blur.
-- **Highlights**: use trimmed segments from the source video as background, with overlay text.
-- **Consistent aesthetic**: all scenes should feel like they belong together — similar color palette, lighting, or style. Specify this in prompts (e.g., "consistent cinematic style, warm tones").
-- **Text legibility**: ensure the background has enough contrast with the subtitle text. If the background is bright, use dark subtitle text or add a semi-transparent overlay.
-
-### BGM — mandatory
-
-Always add a root-level audio node:
-```markdown
-- audio isBackground:true foreground:true src:bgm.mp3 volume:0.1
-```
-
-Choose BGM that matches the style:
-- Cinematic: ambient, orchestral pads, tension builders
-- Energetic: upbeat electronic, lo-fi beats, pop
-- Minimal: soft piano, field recordings, sparse ambient
-- Trendy: viral audio snippets, trending beats
-
-Volume: 0.08–0.15 (lower than vlog BGM — short video TTS should be very clear).
-
-### Effects
-
-- **Hook**: `effect:zoomIn` — dramatic entrance
-- **Body**: `effect:slideInRight` — forward momentum
-- **Body transitions**: `effect:fadeIn` for calm scenes
-- **Key visual reveal**: `effect:bounceIn` for emphasis
-- **CTA**: `effect:zoomIn` or `effect:bounceIn`
-
-### Style-specific rules
-
-| Style | Pace | Voice | Visuals | BGM |
-|---|---|---|---|---|
-| **cinematic** | 4–7s per scene | authoritative, deep | dramatic lighting, nature, architecture | orchestral, ambient |
-| **energetic** | 2–4s per scene | fast, excited, high energy | bright colors, motion blur, urban | upbeat electronic, pop |
-| **minimal** | 5–8s per scene | calm, slow, thoughtful | clean backgrounds, negative space | soft piano, sparse |
-| **trendy** | 2–5s per scene | casual, relatable, meme-aware | dynamic, text-heavy, greenscreen | viral audio, trending |
-
-### Multi-language (variants)
-
-- Same pattern as courseware: `# <lang>` variant block at file end with per-language TTS voice.
-- Subtitle text and script both need `<lang>:"..."` twins.
-
-### Format-specific structure
-
-| Format | Hook angle | Body structure | Visual style |
-|---|---|---|---|
-| **explainer** | "Most people don't know..." | 3–7 facts/steps | metaphorical/abstract backgrounds |
-| **quote** | "This quote changed how I think..." | 1 scene: the quote. Optional: context | minimal, text-focused |
-| **showcase** | "Look at this..." | 3–5 scenes describing aspects of the image | the image itself, zoom/pan details |
-| **highlight** | "The best part was..." | 3–5 key moments from the source | trimmed video segments |
-| **storytime** | "So this happened..." | 3–5 story beats | atmospheric, story-relevant |
-
-## 4. Components & styles
-
-Short videos use markcut's built-in nodes: `image`, `video`, `audio`, `subtitle`, `script`, `effect`. No custom JSX components are required.
-
-### Subtitle configuration (root config line)
-
-```markdown
-# video
-width:1080 height:1920 fps:30 layout:series
-subtitle:{fontSize:"56px",type:"Typewriter",fontFamily:"Arial Black,Impact"}
-```
-
-The subtitle config sets defaults. Per-scene overrides are done via `style:` on the subtitle node.
-
-### Per-scene subtitle override example
-
-For a hook scene that needs a bigger, bouncy caption:
-```markdown
-- subtitle src:"THIS WILL BLOW YOUR MIND" duration:3 type:Bounce
-  fontSize:"72px" style:"color:#ff6b6b;text-shadow:0 4px 20px rgba(0,0,0,.8)"
-```
-
-### Dark overlay (for text legibility on bright backgrounds)
-
-If the background image is bright, add a gradient overlay using an image with `mix-blend-mode`:
-```markdown
-- image isBackground:true src:gradient-overlay.png
-  style:position:absolute;inset:0;width:100%;height:100%;object-fit:fill
-```
-
-Generate a linear-gradient PNG during assembly, or use a CSS gradient in the stylesheet.
-
-### Stylesheet (optional — add if gradients needed)
-
-If you want a built-in dark overlay without a separate image:
-```css
-~~~css stylesheet
-.overlay {
-  position: absolute;
-  inset: 0;
-  background: linear-gradient(180deg, rgba(0,0,0,.1) 0%, rgba(0,0,0,.6) 100%);
-  pointer-events: none;
-}
-~~~
-```
-
-Then use `- component isBackground:true jsx:"<div className='overlay' />"` at root level.
-
-## 5. Workflow
-
-### Phase 0: Understand input
-
-1. **Determine input type** — Is the source a topic/text, an image, a video, or a URL? Read/analyze the source material. For URLs or articles, extract the core message. For images, understand what it shows. For videos, extract key moments or transcribe with STT.
-
-2. **Collect inputs** — Fill §1 inputs table from user conversation. Determine format (explainer/quote/showcase/highlight/storytime) and style (cinematic/energetic/minimal/trendy).
-
-### Phase 1: Script
-
-3. **Write the script** — Fill `prompts/script.md` with the topic, format, and style. This produces a timed scene breakdown:
-   - Hook line + visual idea
-   - Body: 3–7 key points, each with subtitle text + narration + visual prompt
-   - CTA line + visual idea
-   - Total word count within target duration
-
-   **Review the script**: does the hook grab? Is it scannable? Does each scene have one clear idea?
-
-### Phase 2: Storyboard
-
-4. **Generate backgrounds** — For scenes using `src:auto`, run the ITT/TTV CLI for each prompt. For source-media formats (showcase/highlight), extract or prepare the source visuals:
-   - Showcase: source image as-is
-   - Highlight: trim source video segments using `startFrom`/`endAt`
-
-5. **Assemble short.md** — Write the markcut markdown file using §2 grammar. Root config: 1080×1920, `subtitle:{"fontSize":"56px"}`, TTS voice. BGM at root level. Each scene: background media + subtitle + optional script.
-
-### Phase 3: Render & Review
-
-6. **Render** — `npx @lalalic/markcut render short.md`. On errors: fix and re-render, max 3 attempts per error, then ask.
-
-7. **Review (quality gate)** — Run `agents/reviewer.md` in a fresh separate session. Pass: `short.md`, rendered MP4, `TEMPLATE.md`, target duration, language, format, style. Returns `{verdict, findings[]}`.
-
-8. **Fix loop** — On FAIL: fill `prompts/fix.md` with findings, edit, re-render, re-review. Max 3 iterations, then escalate.
-
-## 6. Quality gate — exit criteria
-
-Done only when ALL hold:
-
-- [ ] reviewer verdict = `PASS` (zero blocker/major findings)
-- [ ] total duration within ±15% of target
-- [ ] structure matches §2 (hook → body×N → CTA)
-- [ ] hook ≤ 3 seconds and creates curiosity/surprise (reviewer checks first frame + first subtitle)
-- [ ] no blank/black frames; all `src:auto` prompts generated usable media
-- [ ] STT transcript matches script lines (≥90% content match)
-- [ ] subtitle text is scannable (≤10 words per scene, ≥48px readable at 1080p)
-- [ ] BGM present, ducked under voice (TTS clearly audible over music)
-- [ ] total scene count 5–9 (1 hook + 3–7 body + 1 CTA)
-- [ ] effects applied: hook has `effect:zoomIn` or equivalent, body scenes have entrance transitions
-- [ ] format-specific check:
-  - explainer: every body scene teaches one distinct point
-  - quote: subtitle shows the quote text, script provides context
-  - showcase: source image is visible and well-framed
-  - highlight: source video segments play correctly with startFrom/endAt
-  - storytime: narrative arc with clear beginning, middle, end
-
-## 7. Reference
-
-- Golden example: See `tests/fixtures/templates/courseware.md` for the template format.
-- Markcut subtitle types: `docs/markdown-descriptive.md` §§ "Subtitle" and "Caption animations" for all available `type:` values.
-- Effects reference: `docs/markdown-descriptive.md` for `effect:` values (`fadeIn`, `zoomIn`, `bounceIn`, `slideInLeft`, `slideInRight`, `flipIn`).

package/skills/markcut/docs/templates/short-video/agents/reviewer.md

@@ -1,94 +0,0 @@
----
-name: short-video-reviewer
-description: Read-only quality gate for short-form vertical videos. Checks hook effectiveness, subtitle legibility, pacing, BGM, and format-specific rules.
-context: fresh
-mode: read-only
-tools: read file, bash (ffmpeg/ffprobe, STT CLI), image understanding
----
-
-# System prompt
-
-You are a strict, evidence-based QA reviewer for short-form vertical videos (TikTok/Reels/Shorts). You review; you **never edit** any file.
-
-## Inputs (provided in the task)
-- `short_md` — absolute path to the generated short video markdown
-- `mp4` — absolute path to the rendered video
-- `template_md` — absolute path to this `TEMPLATE.md`
-- `duration_min` — target duration in seconds
-- `language` — expected language
-- `format` — explainer | quote | showcase | highlight | storytime
-- `style` — cinematic | energetic | minimal | trendy
-
-## Procedure
-
-### 1. Static review (source vs rulebook)
-- Structure: hook → body×N → CTA. Hook first, CTA last.
-- Scene count: 5–9 total (1 hook + 3–7 body + 1 CTA).
-- Root config: `width:1080 height:1920` (9:16). If not, blocker.
-- BGM: `- audio isBackground:true` present at root level. If missing, blocker.
-- Each scene has: subtitle text, background image/video, duration. Body scenes have `effect:`.
-- Subtitle text ≤ 10 words per scene. If >10, minor finding.
-- Hook subtitle ≤ 15 words, punchy/curious. If dull, major finding.
-- Format-specific:
-  - explainer: body scenes teach distinct facts.
-  - quote: one scene shows the quote as subtitle, script provides attribution/context.
-  - showcase: source image referenced in src.
-  - highlight: source video with `startFrom`/`endAt` trimming.
-- `src:auto prompt:"..."` present on image/video nodes for explainer/quote/storytime formats.
-
-### 2. Dynamic review
-- `ffprobe`: duration (compare to target ±15%), resolution (should be 1080×1920 or close), audio streams (≥1 for BGM).
-- Extract 6 frames evenly spaced. For each frame:
-  - not blank/black
-  - subtitle text visible and legible (font large enough, contrast good)
-  - background image/video rendered (no broken/placeholder icons)
-- Extract audio → transcribe with STT CLI. Compare against all `script` lines in order. ≥85% match.
-- Spot-check BGM audibility: is the music present but not drowning the TTS?
-
-### 3. Report
-
-Severity: `blocker` (wrong aspect, no BGM, blank frames) · `major` (hook too slow, illegible text, no effects, duration off >20%) · `minor` (font size slightly small, subtitle >10 words, minor timing)
-
-Output exactly this JSON:
-
-```json
-{
-  "verdict": "PASS | FAIL",
-  "measured": {
-    "duration_s": 0,
-    "target_s": 0,
-    "resolution": "",
-    "stt_match_pct": 0,
-    "scene_count": 0
-  },
-  "findings": [
-    {
-      "id": "F1",
-      "severity": "blocker | major | minor",
-      "scene": "<scene name or 'global'>",
-      "check": "<which rule>",
-      "issue": "<what is wrong>",
-      "evidence": "<frame path / STT excerpt / source line>",
-      "fix_hint": "<one-line suggestion>"
-    }
-  ]
-}
-```
-
-`PASS` requires zero `blocker` and zero `major` findings.
-
-# Task template
-
-```
-Review a short-form vertical video against its template and format/style rules.
-
-short_md: {short_md}
-mp4: {mp4}
-template_md: {template_md}
-duration_min: {duration_min}
-language: {language}
-format: {format}
-style: {style}
-
-Follow your procedure. Use whisper-cli (whisper.cpp) for STT if python whisper is unavailable. Save frames under a review directory. Output the verdict JSON only.
-```

package/skills/markcut/docs/templates/short-video/prompts/fix.md

@@ -1,27 +0,0 @@
-# Prompt: Apply Reviewer Findings
-
-> Fill every `{placeholder}`, then execute in your own context.
-
----
-
-Apply the reviewer findings below to `{short_md_path}` with minimal, targeted edits.
-
-Reviewer findings (JSON):
-
-{findings_json}
-
-Rules:
-- Address every `blocker` and `major` finding. Minor findings only when trivially fixable.
-- Hook too slow → shorten hook duration, make text more punchy, use `effect:zoomIn`.
-- Subtitle too small → increase `fontSize` in root config or per-scene.
-- Missing BGM → add `- audio isBackground:true foreground:true src:bgm.mp3 volume:0.1` at root level.
-- Missing effects → add `effect:zoomIn` to hook/CTA, `effect:slideInRight` to body scenes.
-- Blank frames → check `src:auto` prompts generated usable media; if not, add fallback prompts.
-- Wrong aspect → ensure root config is `width:1080 height:1920`.
-
-After editing, list what changed:
-```
-- <finding id>: <scene> — <one-line description>
-```
-
-Then re-render and re-review.

package/skills/markcut/docs/templates/short-video/prompts/script.md

@@ -1,65 +0,0 @@
-# Prompt: Short Video Script
-
-> Fill every `{placeholder}`, then execute in your own context.
-> Produces a timed scene breakdown: hook + body×N + CTA.
-
----
-
-You are writing a {duration}s {format} short video in {style} style.
-
-Topic / source material:
-
-{material}
-
-Source media (if any): {source_media_description}
-
-## Rules
-
-### Hook (scenes[0])
-- One punchy sentence, 5–15 words. Creates curiosity, surprise, or emotional connection.
-- No filler. No introductions. Straight into the hook.
-- Visual idea: one sentence describing an arresting image that matches the hook.
-
-### Body (scenes[1..N])
-- 3–7 scenes. Each 3–8 seconds. One clear idea per scene.
-- Each scene has:
-  - **subtitle**: text shown on screen — ≤10 words, large, scannable
-  - **script**: what TTS says — 5–15 words, conversational, matches the subtitle idea
-  - **visual**: one sentence prompt for the background image (`src:auto prompt:"..."`)
-  - **effect**: entrance animation (`slideInRight`, `zoomIn`, `fadeIn`, `bounceIn`)
-  - **duration**: 3–8 seconds
-- Total body words = {total_words_available} (target_duration × 3)
-
-### CTA (scenes[N+1])
-- One scene, 3–5 seconds.
-- CTA line: call to action (follow, like, share, subscribe).
-- Visual: positive, inviting, memorable.
-
-### Total word budget: {word_budget} words (target duration × 3 w/s)
-
-## Output format
-
-```
-Hook:
-  subtitle: "<hook text>"
-  script: "<optional narration>"
-  visual: "<visual prompt for src:auto>"
-  effect: zoomIn
-  duration: 2
-
-Body 1:
-  subtitle: "<key point 1>"
-  script: "<narration for point 1>"
-  visual: "<visual prompt>"
-  effect: slideInRight
-  duration: 5
-
-Body 2: ...
-
-CTA:
-  subtitle: "<call to action>"
-  script: "<final words>"
-  visual: "<visual prompt>"
-  effect: zoomIn
-  duration: 4
-```