How do I install the Profile Clip skill?

Run npx skillmds@latest add jperrello/profile-clip in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Profile Clip safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Profile Clip?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Profile Clip free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Profile Clip?

jperrello (@jperrello) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Profile Clip by jperrello — Agent Skill

Name: Profile Clip
Author: jperrello

Profile Clip

Sidecar-free style analyzer — reduce ANY finished vertical short (ours in output/ OR a downloaded Group C exemplar) into a versioned Style Profile JSON using ONLY pixels + audio, never our internal edit plans. Deterministic tier (ffmpeg scene-detect for cut rhythm, silencedetect + RMS for speech/SFX/music, MediaPipe face sampling for cutaway fraction, Canny caption-band density) fills every quantitative field; ONE optional Claude vision call on a director-pass contact sheet fills the enumerated judgment fields (broll_mode, caption_style, production_class, hook_device). The shared organ of the style-replication loop — style-corpus profiles Group C exemplars with it and style-gate profiles our span 0 with it, so comparison is fair (delivered pixels vs delivered pixels). Non-fatal, idempotent (.ppmeta). PROFILE_VISION=0 skips the vision call.

jperrello 0ad0e56 5 files · 19.1 KB Updated Sep 17, 2026

File contents

profile-clip

bash .claude/skills/profile-clip/profile-clip.sh <clip.mp4> <out.styleprofile.json> [--no-vision] [--pane <p>]

Output: <out>.styleprofile.json, style_profile_version: 1 per SPEC-style-replication.md — duration_sec, cuts{cuts_per_min, median_shot_sec, p90_shot_sec, longest_static_gap}, speech{words_per_min, max_silence_sec, speech_fraction} (words_per_min via local whisper when available, else absent), visual{face_fraction, cutaway_fraction, cutaway_count, opens_on_face}, captions{present_fraction, band}, audio{onsets_per_min, music_floor_db, music_present}, hook{first_cut_sec, title_overlay_open}, vision{broll_mode, caption_style, production_class, hook_device, confidence}, meta{needs_vision}.

Routing (UniRoute principle): the cheap tier decides everything it can and lists what it can't in meta.needs_vision; exactly one vision round-trip fires only when that list is non-empty and PROFILE_VISION=1. Vision failure → the deterministic profile ships with vision:{} (the match gate only reads deterministic levered fields, so the gate still works).

Env: PROFILE_VISION (1), PROFILE_FRAMES (12), PROFILE_SAMPLES (16 sampled frames for face/caption stats), PROFILE_SCENE (0.3), PROFILE_TRANSCRIPT (pre-existing transcript json to skip whisper), PROFILE_MUSIC_DB (-48).

---name: profile-clipdescription: Sidecar-free style analyzer — reduce ANY finished vertical short (ours in output/ OR a downloaded Group C exemplar) into a versioned Style Profile JSON using ONLY pixels + audio, never our internal edit plans. Deterministic tier (ffmpeg scene-detect for cut rhythm, silencedetect + RMS for speech/SFX/music, MediaPipe face sampling for cutaway fraction, Canny caption-band density) fills every quantitative field; ONE optional Claude vision call on a director-pass contact sheet fills the enumerated judgment fields (broll_mode, caption_style, production_class, hook_device). The shared organ of the style-replication loop — style-corpus profiles Group C exemplars with it and style-gate profiles our span 0 with it, so comparison is fair (delivered pixels vs delivered pixels). Non-fatal, idempotent (.ppmeta). PROFILE_VISION=0 skips the vision call.---# profile-clip```bashbash .claude/skills/profile-clip/profile-clip.sh <clip.mp4> <out.styleprofile.json> [--no-vision] [--pane <p>]```Output: `<out>.styleprofile.json`, `style_profile_version: 1` per`SPEC-style-replication.md` — `duration_sec`, `cuts{cuts_per_min, median_shot_sec,p90_shot_sec, longest_static_gap}`, `speech{words_per_min, max_silence_sec,speech_fraction}` (words_per_min via local whisper when available, else absent),`visual{face_fraction, cutaway_fraction, cutaway_count, opens_on_face}`,`captions{present_fraction, band}`, `audio{onsets_per_min, music_floor_db,music_present}`, `hook{first_cut_sec, title_overlay_open}`, `vision{broll_mode,caption_style, production_class, hook_device, confidence}`, `meta{needs_vision}`.Routing (UniRoute principle): the cheap tier decides everything it can and listswhat it can't in `meta.needs_vision`; exactly one vision round-trip fires onlywhen that list is non-empty and `PROFILE_VISION=1`. Vision failure → thedeterministic profile ships with `vision:{}` (the match gate only readsdeterministic levered fields, so the gate still works).Env: `PROFILE_VISION` (1), `PROFILE_FRAMES` (12), `PROFILE_SAMPLES` (16 sampledframes for face/caption stats), `PROFILE_SCENE` (0.3), `PROFILE_TRANSCRIPT`(pre-existing transcript json to skip whisper), `PROFILE_MUSIC_DB` (-48).

Frequently asked questions