agentleFS
duncan-buildroomduncan-buildroom/profit-room-skills/skills/yap/SKILL.md

yap

Skill18 starsChanged 4 months agoMITView on GitHub

Create a viral "Yap-style" short-form video for Instagram Reels / TikTok / YouTube Shorts — researches a trending topic in your niche, drafts an engagement-engineered talking-head script using the proven Yap format (contrarian hook + four-part structure), then renders it as a HeyGen avatar speaking the script (9:16, burned-in captions). Use when the user says /yap, "make a yap video", "create a yap-style reel", "yap script + avatar", "talking-head reel about X", or wants an AI-avatar short-form video from a trending topic.

4 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

10 rules. 0 are common, 8 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If it doesn't exist, run the personalize skill first (or just tell me your name, niche, offer, and audience — I'll create it).

    line 17
  2. Distinctive:

    If a field this skill needs is blank, I'll ask one or two quick questions and save the answers back to the profile so you only answer once.

    line 18
  3. Distinctive:

    Never invent facts or results about the user — use only what's in the profile, or ask.

    line 19
  4. Very specific:

    Read ~/.claude/brand-profile.md so the topic + voice are accurate to *your* audience.

    line 46
  5. Distinctive:

    If a needed field is blank, ask one quick question and save the answer back to the profile.

    line 53
  6. Distinctive:

    If the user named a topic, skip discovery and use it.

    line 61
  7. Very specific:

    Use firecrawl_search (preferred — load via ToolSearch select:mcp__firecrawl__firecrawl_search) or WebSearch.

    line 63
  8. Distinctive:

    Write to be saved (reference value) or sent (relatable/contrarian) — not just liked.

    line 97
  9. Distinctive:

    Keep the spoken script under ~110 words for a 30s cut unless the user asks for longer (HeyGen will speak everything — long scripts = long …

    line 139
  10. Distinctive:

    Never go full-faceless-spam — this is the user's avatar speaking their real insight.

    line 142

“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Yap — Viral Talking-Head Reel Generator

Turns a trending niche topic into a finished AI-avatar Reel: research → script → HeyGen render.

"Yap-style" = direct-to-camera, fast-paced spoken monologue/rant/storytime, high energy, burned-in karaoke captions. Follow the format rules below exactly — they are what make the format work.


Personalization

This skill works in your brand voice. Before running, load the brand profile at ~/.claude/brand-profile.md.

  • If it doesn't exist, run the personalize skill first (or just tell me your name, niche, offer, and audience — I'll create it).
  • If a field this skill needs is blank, I'll ask one or two quick questions and save the answers back to the profile so you only answer once.
  • Never invent facts or results about the user — use only what's in the profile, or ask.

Requirements

  • HeyGen — a HeyGen account with at least one avatar. Connect the HeyGen MCP (exposes create_video_from_avatar + get_video) or use the HeyGen API directly. Never hard-code an API key; read it from the environment.
  • Your HeyGen avatar ID — from your brand profile field HeyGen-avatar-id, or set HEYGEN_AVATAR_ID. If neither is set, list your HeyGen avatars and ask the user which one to use, then save it back to the profile.
  • (Optional) Your HeyGen voice ID — from brand profile field HeyGen-voice-id, or set HEYGEN_VOICE_ID. If neither is set, omit it and let the avatar's default voice speak.
  • Research tools (any one): firecrawl_search MCP, or built-in WebSearch. If none are available and the user didn't name a topic, ask them for one.
  • Fallback — if HeyGen isn't connected, still write and deliver the script, then tell the user to connect HeyGen to render.

HeyGen MCP tools may be deferred/namespaced under your HeyGen connector. Load them first, e.g.: ToolSearch({ query: "select:create_video_from_avatar,get_video", max_results: 5 })


Locked render configuration (do not ask, use these)

  • Aspect ratio: 9:16 · Resolution: 1080p · Captions: burned-in (style: "default")
  • Voice speed: 1.1 (yap energy — fast but intelligible)
  • Avatar / voice: pulled from your profile / env (see Requirements). Never hard-code an avatar or voice ID in this file.

Step 1 — Load your niche context

Read ~/.claude/brand-profile.md so the topic + voice are accurate to your audience. Pull:

  • Your audience — who you're talking to and their pain points.
  • Your offer — what you sell / the system you teach (for CTA context only).
  • Your voice rules — how you sound, and what to avoid.
  • Recent hooks — so you don't repeat one you've already used (if the profile tracks them).

If a needed field is blank, ask one quick question and save the answer back to the profile.

Your niche comes from the profile. If it's not set, ask: "What's your niche / what do you make content about?" and save it.


Step 2 — Research trending topics

If the user named a topic, skip discovery and use it. Otherwise find what's hot right now.

Use firecrawl_search (preferred — load via ToolSearch select:mcp__firecrawl__firecrawl_search) or WebSearch. Run 3–4 searches in parallel, biased to the last 2–4 weeks, templated to the user's niche. Example shapes (swap in the profile's niche and audience):

  • new [niche] tool launch [current month year] [audience]
  • [niche] trending reddit [current month]
  • [niche] debate / controversy [current year]
  • [niche] news creators this week

Pick ONE topic that scores highest on:

  1. DM-shareable — would someone send this to a friend? (the #1 reach signal)
  2. Contrarian angle available — is there a "everyone thinks X, actually Y"?
  3. Maps to audience pain — connects to a real frustration from the profile.

State the chosen topic and the angle in one line before scripting.


Step 3 — Draft the Yap script

Target 30 seconds ≈ 70–110 words, fast delivery. Follow the four-part structure exactly:

BeatTimeWordsJob
Hook0–3s≤12Stop scroll + state the promise. NO "hey guys".
Body3–24s~70Deliver. One idea per sentence. Vary length: punch. then longer. punch.
Payoff24–28s~15The actual answer, plainly. Don't withhold it.
Loop/CTA28–30sshortLoop to the hook OR a question that demands a comment.

Hook — default to the CONTRARIAN (highest-verified) formula:

"Everyone says [common belief], but [contrarian truth]."

Other allowed openers: curiosity gap, callout ("If you're [X], stop doing [Y]"), bold number/result. Always deliver the promise by the 3-second mark.

Engagement engineering (non-negotiable):

  • Open a loop early ("and it's not what you think…"), close it at the payoff.
  • Write to be saved (reference value) or sent (relatable/contrarian) — not just liked.
  • End on a line that earns a DM or a comment.
  • Plain, spoken, first-person. Match the profile's voice rules. No corporate filler, no throat-clearing.

Write the whole thing in your tone and from your take — the opinion is yours, grounded only in what's in the profile. Output the script as clean speakable text (this is what HeyGen speaks — no stage directions, no emojis, no markdown inside the script string). Show the user the script and the chosen hook, then proceed to render (don't wait for approval unless they ask).


Step 4 — Render the HeyGen avatar video

Resolve the avatar ID (profile field HeyGen-avatar-id → env HEYGEN_AVATAR_ID → ask). Resolve the voice ID the same way (optional; omit if unset).

Call create_video_from_avatar:

avatarId:    <your resolved HeyGen avatar ID>
voiceId:     <your resolved HeyGen voice ID, or omit for the avatar default>
script:      <the script text from Step 3>
aspectRatio: "9:16"
resolution:  "1080p"
caption:     { style: "default" }      // burns karaoke captions in + returns SRT sidecar
voiceSettings: { speed: 1.1 }
title:       "Yap — <topic slug>"

Capture the returned video_id.


Step 5 — Poll & deliver

Poll get_video({ videoId }) every ~20s until status is completed (or failed). Don't spam — space the calls.

  • On completed: give the user the video_url, duration, and the subtitle_url (SRT).
  • On failed: report the failure reason and offer to re-render (e.g. shorten script if it errored on length).

Save the final script + topic + video_url to ~/yap-outputs/<date>-<slug>.md so there's a record.


Guardrails

  • Keep the spoken script under ~110 words for a 30s cut unless the user asks for longer (HeyGen will speak everything — long scripts = long videos).
  • The script string must be plain speech only. No emojis, no "[pause]", no markdown.
  • Funnel level is REACH (cold discovery): one value hit, soft engagement CTA. No product pitch or links unless the user asks (and if they do, use their offer from the profile).
  • Never go full-faceless-spam — this is the user's avatar speaking their real insight. Value first.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.