whos-wrong
Skill2 starsChanged 15 days agoMITView on GitHub
Make or fix a "Who's Wrong?" episode: a petty everyday argument where both sides have a point, played fast by Bram and Tess, with the verdict meter tipping as each point lands and a comment ask at the end. Covers the premise rules, the meter, the argument style, wording, staging, render and hand-back. Use when asked for a Who's Wrong episode, an AITA-style skit, a "who's right" debate short, or to restage split-evenly. The team writes the twist; never invent one and present it as final.
In this repository13 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- crew-episodeSkill
- explainerSkill
- science-shortSkill
- shorts-editorSkill
- skit-directorSkill
- story-episodeSkill
- the-catch-stagingSkill
- the-catch-storySkill
- toon-characterSkill
- topic-to-videoSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
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
- Very specific: line 14
Read first: src/data/series/whos-wrong.json (the bible), src/data/characters/bram.md and tess.md, docs/skits.md (verdict meter, argument …
- Distinctive: line 23
Both sides defensible: Before staging, write one sentence for each side ("Bram: friends split evenly, it's simpler".
- Distinctive: line 26
Invented situations only: Never retell a real Reddit post or a real person's story.
- Very specific: line 43
No narrator, no "fun fact", no moral, no sign-off.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Who's Wrong?
An episode is an 18–28 s argument about something small that everyone has an opinion on: the
bill, the armrest, the last fry, the group chat. Both sides have a real point. The verdict meter
(overlay.verdict) stamps WHO'S WRONG? on frame 0 and tips toward whoever sounds worse as the
fight goes. The last line flips sympathy, and the episode ends on "Comment: BRAM or TESS?".
The engagement is the comments: if the split isn't real, nobody argues and the short dies.
Read first: src/data/series/whos-wrong.json (the bible), src/data/characters/bram.md and
tess.md, docs/skits.md (verdict meter, argument style, cut-off lines, wording) and
skills/skit-director/SKILL.md (general staging). Reference episode: public/skits/split-evenly/.
Channel plan: ../planning/2026-10-29-new-viral-channel-idea/discussion.md.
Hard rules
- The team writes the twist (the last line that flips sympathy). You may draft one to render a sample, but say in the hand-back that it is a placeholder.
- Both sides defensible. Before staging, write one sentence for each side ("Bram: friends split evenly, it's simpler"; "Tess: she had a salad"). If you can't, the premise is wrong. The meter should finish between 40 and 60.
- Invented situations only. Never retell a real Reddit post or a real person's story.
- Out of bounds: real people, men-vs-women framing ("women always…"), medical or legal advice. Petty is the point; mean is not. Swap who started it from one episode to the next.
Script shape
| Time | Beat | Meter |
| 0–0.3 s | The first line is already the fight ("Sixty-two dollars. Let's split it evenly.") | 50 → slight lean |
| to ~15 s | Point and counterpoint, each line shorter than the last; one cut-off ("—") | swings 30 ↔ 80 |
| ~15–20 s | The setup for the twist: an innocent question ("What was last Friday?") | holds |
| last line | The twist (team-written), punchline: true, a slam on the key word | back to 40–60 |
| +1.5 s | Reaction close-up + the comment ask (automatic) | — |
- 8–12 lines. Lines under 10 words; never over 14 (the check warns).
- Contractions, fragments, interruptions. Each has a habit: Bram restates his made-up rule, Tess quotes the receipt.
- No narrator, no "fun fact", no moral, no sign-off. The wording check lists phrases that read
model-written; fix every
ai-wordingwarning. verdictgoes on the lines that score a point, not on every line.
Skit
{
"schemaVersion": 2,
"meta": { "title": "…", "description": "<the question> Who's wrong?", "hashtags": ["whoswrong", "aita", "stickfigure"], "series": { "id": "whos-wrong", "season": 1, "episode": 2 } },
"set": "cafe-1", // homeSets in the series; pick by the situation
"style": "argument", // the series sets it too; say it anyway
"coldOpen": "none", // the stamp is the hook
"cast": [
{ "id": "bram", "character": "bram", "mark": "left" }, // left = the meter's red side
{ "id": "tess", "character": "tess", "mark": "right" }
],
"overlay": { "verdict": {}, "subtitles": true },
"beats": [
{ "id": "l1", "speaker": "bram", "line": "…", "verdict": 45 }
]
}Swap the marks (and so the meter's sides) when Tess starts it. The meter's left side is the
first cast member unless sides says otherwise.
Steps
pnpm voice:tts <id> # Kokoro, fast pace from the argument style (or the harness voices it)
pnpm check <id> # self-check: dead air, speech rate, first word, line length, wording
pnpm render <id> out/<id>.mp4Then watch it as video (not stills): the stamp must not cover a face, the meter should tip visibly on each point, and the ask must be readable for at least 1.5 s.
Hand-back
- The video path, the length, and the self-check result (paste any warnings).
- One sentence per side, and where the meter ends.
- Whether the twist is the team's or your placeholder.
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
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.

