audit-short-1v1
Skill1 starsChanged 5 months agoMITView on GitHub
Audit pré-publication YouTube d'un short {{GAME_PROJECT}} 1v1. Score viralité 0-100 + verdict GO/WARN/NO-GO + recommandations basées sur algo YouTube Shorts 2026 + forensic du channel {{YT_CHANNEL}} (5 métriques pondérées : durée 35%, iconic personas 30%, action density 20%, hook variance 15%, bonus winner overlay). Use when about to upload a 1v1 short, or when user asks "audit ce short" / "est-ce que cette vidéo va marcher" / "score viralité" pour un .mp4 1v1.
In this repository11 other files it gives its agents.
- anti-ai-detect-writingSkill
- audit-short-ffaSkill
- awesome-design-mdSkill
- claude-councilSkill
- deep-researchSkill
- design-mdSkill
- design-audit-paris-nuitSkill
- gemini-deep-researchSkill
- gemini-spriteSkill
- oracle-ampere-a1-setupSkill
- smart-routerSkill
The file
Files
Skill : audit-short-1v1
Audit pré-publication d'un short {{GAME_PROJECT}} format 1v1. Score 0-100 + verdict GO/WARN/NO-GO + recommandations actionnables.
Prérequis
cd ~/Desktop/battle-engine
source .venv/bin/activate
# Deps : cv2 (deja), imagehash, pytesseract (deja installés)
# Tesseract binary : /opt/homebrew/bin/tesseract (deja present)Invoquer
cd ~/Desktop/battle-engine
source .venv/bin/activate
python3 pipeline/audit_cli.py --preset 1v1 /path/to/video.mp4
# Avec matchup explicite (optionnel, sinon parsé du nom de fichier) :
python3 pipeline/audit_cli.py --preset 1v1 /path/to/video.mp4 --matchup mario,pikachuExit code : 0 si GO ou WARN, 1 si NO-GO.
Métriques et poids
| Métrique | Poids | PASS | WARN | FAIL |
duration_s | 35% | 25-45s | 18-24s ou 46-60s | <18s ou >60s |
iconic_score | 30% | ≥2 icônes | 1 icône | 0 icône |
action_density | 20% | ≥14 | 8-13 | <8 |
hook_variance_t2 | 15% | ≤734 | 735-1200 | >1200 |
Icônes : Mario, Pikachu, Sonic, Goku, Steve (Minecraft).
Bonus : +5 pts si overlay winner détecté (OCR sur les 5 dernières frames).
Hard gates NO-GO : résolution != 1080x1920, black_frame_ratio > 10%, freeze_ratio > 15%, durée > 90s.
Interpréter le score
| Score | Verdict | Action |
| 60-100 | GO — top tier probable | Uploader |
| 40-59 | WARN — mid tier | Envisager re-render ou uploader quand même |
| 0-39 | NO-GO — sous-performance attendue | Re-render avec autre seed ou changer matchup |
Floor 55 : si une métrique est critique (FAIL avec poids élevé), le score est capé à 55 même si les autres sont PASS.
Que faire si NO-GO
- duration_s FAIL (>60s) : Re-render avec autre seed. Réduire HP de 20% pour finish plus rapide.
- iconic_score FAIL (0) : Remplacer 1 perso par Mario/Pikachu/Sonic/Goku/Steve. +67% vues en moyenne sur le channel.
- action_density FAIL (<8) : Combat statique. Vérifier hitboxes ou augmenter agressivité IA. Re-render.
- hook_variance_t2 FAIL (>1200) : Intro corrompue ou bug de spawn à t=1.5s. Re-render impératif.
- freeze_ratio FAIL : Frames figées. Bug Godot probable. Re-render.
Exemple output JSON
{
"score": 100,
"verdict": "GO — top tier probable",
"preset": "1v1",
"hard_gate": null,
"floor_triggered": false,
"metrics": {
"duration_s": 35.57,
"resolution": "1080x1920",
"hook_variance_t2": 0.0,
"action_density": 20.57,
"black_frame_ratio": 0.0,
"freeze_ratio": 0.0,
"winner_text_present": true,
"multi_fighter_visible_t2": 5,
"iconic_score": 2,
"matchup": ["mario", "pikachu"]
},
"contributions": {
"duration_s": {"value": 35.57, "verdict": "PASS", "contribution": 35.0},
"iconic_score": {"value": 2.0, "verdict": "PASS", "contribution": 30.0},
"action_density": {"value": 20.57, "verdict": "PASS", "contribution": 20.0},
"hook_variance_t2": {"value": 0.0, "verdict": "PASS", "contribution": 15.0}
},
"bonus": {"winner_overlay": "+5"},
"recommendations": []
}Calibration (24/05/2026)
Calibré sur 111 vidéos {{YT_CHANNEL}}.
- Seuils forensic (Pearson) : hook_variance -0.734, duration -0.513, action_density +0.184, iconic +0.264.
- Sweet spot duration 1v1 : 27-35s (batch 18-20 successful).
- Personages iconiques both = avg 723 views, none = 477 views.
OUTPUT CONTRACT
Respecte ~/.claude/rules/output-contract.md.
Attendu : JSON output stdout (format exemple ci-dessus) + exit code 0/1 selon verdict
Checks :
- JSON sortie parseable :
python3 -c "import json,sys; json.loads(sys.stdin.read())"≤ exit 0 - Champs obligatoires :
score,verdict,preset,metrics,contributions,recommendations score∈ [0, 100]verdict∈ {"GO — top tier probable", "WARN — mid tier", "NO-GO — sous-performance attendue, refaire"}- Exit code 0 si score ≥ 40, exit 1 si < 40
- Si NO-GO :
recommendationsnon vide avec action concrète par métrique FAIL - Hard gates checkés (résolution, freeze, black, durée)
Format réponse : LIVRABLE: audit 1v1 / SCORE: X / VERDICT: GO|WARN|NO-GO / METRICS: dur=A iconic=B action=C var=D / RECO: <liste si NO-GO> / NEXT: <upload | re-render seed Y | changer matchup>
Anti-hallucination : Si le script Python a échoué : INCOMPLETE: audit_cli.py failed (<erreur>). Si la vidéo n'est pas 1080×1920 : hard gate NO-GO immédiat.
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.

