Skip to content
Recaply.
Sign inSubmit a tool

DirectorySkills

Full size
Run 1Inputs to a result, then proof“In talk.mp4 merge the three things we tried into "40% less meeting time", then show a quote.”
Runs in
  • Claude Code
  • Codex
  • Cursor
  • Gemini CLI
  • GitHub Copilot
  • OpenCode
Generates with
Your agent's model
Tested with
Claude Code 2.1.322Claude Sonnet 5.51 real run, 9 Nov
Needs
Node 22+Chromeffmpeg
Licence
Recaply licence

Inputs to a Result, Then Proof

Inputs merge into one result, then a quote card backs it up.

$9.99once, yours to keep

Every premium skill and prompt, one plan$19.99 a month. Cancel any time. What you get when you buy

Payments are non-refundable. See our Terms of service.

Adds a merge diagram to a talking-head video (inputs combining into one result) and then a quote that backs up the result, handing over in the same panel. Timed to the words they say, on their own footage, with their sound kept; 16:9, 9:16, 1:1 or 4:5. Renders locally with Chrome and ffmpeg: no video model, no key and no cost. Use when someone explains what led to a finding and then backs it with a quote or source.

By RecaplyUpdated 11 Nov

Install Inputs to a Result, Then Proof

After the download, ask your agent to add it, or run the line yourself.

Read https://recaply.co/skills/merge-to-evidence-overlay/install.md and follow it to install the merge-to-evidence-overlay skill from merge-to-evidence-overlay.zip.

Paste it into your agent. It reads our install steps, finds the ZIP you downloaded and adds the skill for you.

New to skills? How to install a skill

Similar skills

All 735 skills

Who Inputs to a Result, Then Proof is for

What you get

  • Your video with the overlay timed to the words, sound kept
  • The inputs land and join into one result in the accent; then the panel hands over to a quote that proves the point
  • 16:9, 9:16, 1:1 or 4:5

Good fit

  • When someone explains what led to a finding and then backs it with a quote or source

Not for

  • Footage with nobody speaking
  • Cutting or re-editing the video itself

About this skill

You say

In talk.mp4 merge the three things we tried into "40% less meeting time", then show a quote.

It said back

I've made the video: talk-graphics.mp4 in the work folder. It is 1920x1080, 24.3 s, 30 fps, and the sound is the speaker's own and unchanged. I checked the layout and looked at one frame from each beat. No warnings or notes came up, but I haven't watched the finished render.

  • Merge (11.2–21.2 s): three boxes join into "40% less meeting time". The speaker slides to one side and the diagram fills the other half.
  • "Written updates on Friday" appears as she says it.
  • "One shared board" and "15-minute call when stuck" appear as she says each.
  • The result appears when she says "dropped by 40%".
  • Quote (22.1–24.3 s): "Nobody missed the old Monday call." appears as she says "nobody missed the old", with "Nobody missed" underlined.
Show more

What it did in run 1

  1. merge-to-evidence-overlay
  2. cd .; ls -la; S=.claude/skills/merge-to-evidence-overlay/scripts; node $S/check.mjs
  3. ls -la; node .claude/skills/merge-to-evidence-overlay/scripts/check.mjs
  4. echo '{"video":"talk.mp4","transcript":"talk.json"}' > overlay.json; node .claude/skills/merge-to-evidence-overlay/scripts/overlay.mjs overlay.json --words
  5. cat > overlay.json <<'E' { "video": "talk.mp4", "transcript": "talk.json", "look": "paper", "beats": [ { "beat": "flow", "cue": "Written updates on Friday", "layout": "merge", "place": "side", "items": ["Written updates on Friday", "One shared board", "15-minute call when stuck"], "cues": ["Written updates", "one shared board", "15-minute call"], "result": "40% less meeting time", "until": "dropped by 40%" }, { "beat": "quote", "cue": "nobody missed the old", "place": "side", "text": "Nobody missed the old Monday call.", "by": "The speaker", "highlight": ["Nobody missed"], "for": 3 } ] } E node .claude/skills/merge-to-evidence-overlay/scripts/overlay.mjs overlay.json --face; node .claude/skills/merge-to-evidence-overlay/scripts/overlay.mjs overlay.json --check
  6. ./face-check.png
  7. ./overlay.json
  8. node .claude/skills/merge-to-evidence-overlay/scripts/overlay.mjs overlay.json --face
  9. ./face-check.png
  10. ./overlay.json
  11. node .claude/skills/merge-to-evidence-overlay/scripts/overlay.mjs overlay.json --check
  12. ./overlay-check.png
  13. node .claude/skills/merge-to-evidence-overlay/scripts/overlay.mjs overlay.json 2>&1 | tail -8
  14. Done

9 Nov 2026 · Claude Code 2.1.322 with Claude Sonnet 5.5 · 1 min 39 s · $0.12 of API use

Questions

Is Inputs to a Result, Then Proof free?

No. Inputs to a Result, Then Proof costs $9.99 once and stays in your Recaply account, or it comes with the Recaply plan ($19.99 a month), which opens every premium skill and prompt.

Which agents can use Inputs to a Result, Then Proof?

Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot and OpenCode.

What do I get when I buy Inputs to a Result, Then Proof?

A ZIP of the whole skill folder (106 KB): the SKILL.md your agent follows and its 12 other files, the references, templates and scripts it uses. You download it from this page whenever you need it, and install it in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot and OpenCode. The Recaply plan opens it too, with every other premium skill and prompt.

How do I install Inputs to a Result, Then Proof?

Buy it, then download the ZIP from this page. For Claude Code, unzip it into .claude/skills/ in your project. For Codex, Cursor and Gemini CLI, unzip it and run npx skills add ./merge-to-evidence-overlay with your agent's flag.

What do I get from Inputs to a Result, Then Proof?

Your video with the overlay timed to the words, sound kept. The inputs land and join into one result in the accent; then the panel hands over to a quote that proves the point. 16:9, 9:16, 1:1 or 4:5.

What does Inputs to a Result, Then Proof need?

Node 22+, Chrome and ffmpeg, and no API key. Each use costs nothing beyond your agent.

How big is Inputs to a Result, Then Proof?

13 files, about 80,109 tokens in all. The SKILL.md is about 2,446 tokens, and your agent reads it only when the job fits.

What does a run cost?

Our 1 runs took 1 min 39 s to 1 min 39 s and $0.12 to $0.12 of API use each, in Claude Code with claude-sonnet-5-5. On a Claude plan, a run counts toward your usage instead.

Do I have to name Inputs to a Result, Then Proof to use it?

Type /merge-to-evidence-overlay first to make sure it loads. Without it, Claude Code may judge a small job simple enough to do without the skill.