← ClaudeAtlas

viral-checklisted

Audit a finished short-form video (TikTok / Reels / Shorts) against a virality rubric and ITERATE until it passes — see it, critique it, fix it, re-render, re-check. Combines an objective scan (format, loudness/LUFS, pacing/cut-rate, audio hook accent via scripts/ analyze.mjs) with visual judgment from reading key frames (hook strength, caption legibility & sync, safe areas, B-roll relevance, transitions, ending/loop, trend fit for LatAm/Peru). Use this whenever the user asks to check/verify/QA whether a video is viral or ready to post, score a reel's virality, improve hook/retention, "make sure this goes viral", or iterate on a clip until it's good enough to publish — even if they don't name the rubric. Pairs with the hyperframes-video skill, which builds and edits the actual composition.
GiomarDev/ReelForge · ★ 0 · Code & Development · score 70
Install: claude install-skill GiomarDev/ReelForge
# viral-check — audit a short-form video for virality and iterate to pass This skill is a **reviewer with a feedback loop**, not an editor. It judges a rendered clip against what actually makes short-form video perform, tells you exactly what's failing, drives the fix (usually back through the `hyperframes-video` skill), re-renders, and re-checks — looping until the clip clears the bar or you hit the iteration cap. > A clip is **"viral-ready"** only when the objective scan has **no fails** AND every visual > item in the rubric holds. "Viral" can never be guaranteed (distribution is luck + timing) — > what this enforces is that the video has *no avoidable reasons to flop*. ## The loop ``` render ─▶ analyze.mjs (objective) ─▶ read key frames (visual) ─▶ score vs rubric ▲ │ └────────────── apply concrete fixes, re-render ◀── any FAIL? ────┘ (stop at all-pass, or after ~4 iterations) ``` ### 1. Objective scan ```bash node scripts/analyze.mjs path/to/video.mp4 # human-readable node scripts/analyze.mjs path/to/video.mp4 --json # machine-readable scorecard ``` It reports, with pass/△warn/✗fail thresholds: vertical 9:16, resolution, duration, fps, audio present, **integrated loudness (LUFS)**, true peak, **pacing (cuts/min)**, and a **hook audio accent** (is there a riser/whoosh/impact in the first 1.5s?). Treat every `✗` as a blocker. ### 2. Visual judgment (read the