write-copy

name	write-copy
description	Write the words — scroll-stopping hooks, framework-driven scripts, and platform-ready captions with hashtags. Use when the user needs hook ideas, a video/ad script, ad copy, post captions, or says "what should I post", "write a script", "give me hooks", "write the caption".
allowed-tools	Bash Read Write Edit Glob Grep

Write Copy (hooks · script · caption)

The creative bottleneck isn't generating media — it's knowing what to say. This skill generates hooks, scripts, and captions using proven direct-response frameworks, then lets the user pick and refine. Backed by workflow commands — call the CLI, don't write a script:

node workflows/cli.cjs generateHooks '<json-args>'     # also: generateScript, generateCaption

The TypeScript call signatures shown below document each command's arguments — translate them into node workflows/cli.cjs <command> '<json-args>', don't paste them into a new file.

Pull context from the project first when it exists (project.md audience/pain points/offer; brand.md tone/restrictions) so the copy is on-brand and never violates restrictions.

---

STEP 0: Check for reference copy BEFORE writing anything

Voice consistency comes from examples, not adjectives. Before generating hooks, scripts, or captions:

1. Check the registry for saved copy examples:

   node workflows/cli.cjs resolveAsset '["{name}","style-copy"]'   # a style_references id holding copy examples
   cat projects/{name}/assets/style/copy-examples.md 2>/dev/null
   cat projects/{name}/style-samples/caption-examples/*.md 2>/dev/null   # user's drop inbox
   

   Found → extract the patterns (sentence length, emoji use, hashtag style, CTA phrasing, formality) and MIRROR them in everything you write below.
2. Nothing saved → ask once:

   "Do you have past posts, captions, or scripts that worked well — or accounts whose copy style you like? Paste 1-3 examples (or 'none' — I'll go from brand.md)."

   • Examples given → save to projects/{name}/assets/style/copy-examples.md and register (registerAsset '["{name}","style_references",...]') so future pieces reuse them.
   • "none" → proceed from brand.md tone and state which tone assumptions you made.
3. When examples exist, note in prompts.txt which reference the copy was matched to.

---

STEP 1: Hooks (pick the angle before writing the whole thing)

Generate a spread of hooks across angles and let the user choose 1–2 to build on.

node workflows/cli.cjs generateHooks '{"topic":"AI study planner app","audience":"uni students prepping for finals","painPoint":"no idea where to start revising","platform":"tiktok","count":8}'
# h.data.hooks: [{ text, angle }]  — angles: problem | curiosity | result | contrarian | question | stat | story

Show them grouped by angle and ask which to use. Don't auto-pick — the hook decides the whole piece.

---

STEP 2: Script (framework-driven, sized to duration)

Once a hook is chosen, write the script with a framework that fits the goal:

Framework	Best for
hook-retain-reward	short-form video (TikTok/Reels/Shorts) — default
PAS (Problem-Agitate-Solution)	pain-led ads
AIDA (Attention-Interest-Desire-Action)	classic direct response
BAB (Before-After-Bridge)	transformation / results
star-story-solution	testimonials / founder story

# hook: pass the chosen hook text from generateHooks
node workflows/cli.cjs generateScript '{"brief":"AI study planner that builds a revision schedule in 60 seconds","framework":"PAS","platform":"tiktok","durationSeconds":30,"hook":"<chosen hook text>","cta":"Try it free — link in bio"}'
# s.data: { framework, sections:[{label, voiceover, visual}], wordCount, estimatedDurationSeconds }

Show the script as labelled sections (voiceover + visual cue) and the estimated duration. Flag if estimatedDurationSeconds is far from the target so the user can trim/expand. This voiceover text feeds straight into generateVoiceover() and the per-scene visual cues feed generateStoryboard() / the video skill.

---

STEP 3: Caption (publish-ready, per platform)

The last mile — the post copy + hashtags that actually ship with the video/image.

node workflows/cli.cjs generateCaption '{"topic":"AI study planner app","platform":"instagram","tone":"friendly, confident","cta":"Link in bio","count":3}'   # platform: tiktok | instagram | youtube | facebook
# c.data.captions: [{ text, hashtags:[] }]

Generate per-platform variants if the piece is cross-posted (each platform has different length + hashtag norms). Save the chosen caption next to the output, e.g. projects/{name}/output-contents/.../caption.md.

---

How it connects

generateHooks → pick → generateScript → { voiceover → generateVoiceover,
                                          visual cues → generateStoryboard → video }
                                       → generateCaption → caption.md

Always suggest and confirm — present options, recommend one with a reason, let the user edit. Never fabricate stats/testimonials (pull only real ones from project.md).