q00/openclip · Archived

oc-hook-finder

Internal OpenClip worker. Invoke only when dispatched by the public `oc` skill; do not use this role as the user-facing entry point. Scans ONE transcript section for the most "hooky" moments — surprising claims, strong payoffs, emotional beats, crisp Q&A — and returns ranked short candidates with clean in/out points. Fanned out across sections in parallel for flow 3 (turn a longform's hooks into shorts).

First seen Jul 6, 2026

Installation

$ npx skills add q00/openclip --skill oc-hook-finder

Stronger alternatives

This repository is archived — consider an actively maintained alternative.

Similar popular skills

Related neighbors and high-traction skills in the same topics — useful to compare before installing.

Also in this package

Other skills from q00/openclip · top by installs.

npx skills add q00/openclip

Browse all from q00/openclip

More details

Agent compatibility

Declared targets from SKILL.md / docs. Unmarked agents are not listed — the skill may still install via the CLI.

Claude Code Not declared
Cursor Not declared
Codex Not declared
GitHub Copilot Not declared
Windsurf Not declared
Gemini CLI Not declared
Cline Not declared
OpenCode Not declared

Repository health

Stars 2
License LICENSE
Default branch main
Open issues 0
Status Archived

Package contents

Files included with this skill beyond the listing page.

  • skill md SKILL.md 2,200 B
  • docs SUMMARY.md 433 B

History

  1. First seen on skills.sh
  2. First recorded snapshot · 4 installs

SKILL.md

Hook Finder

Find the moments worth clipping into a short. You own one section; many of you run in parallel across the timeline.

Inputs

  • PROJECT, the section start/end. Read candidates from

<PROJECT>/transcript.json (segments in range). Use analysis.json scene/ silence points to pick clean boundaries when available.

  • Honor any STEERING directive in your assignment (e.g. "only hooks about

pricing", "keep the lecturer's face in frame") over your default ranking.

What makes a hook

  • A standalone payoff (makes sense with zero prior context).
  • Surprise / strong opinion / a number / a vivid example / a punchy Q→A.
  • Starts on a strong line (not a continuation like "and so…", "which means…") and ends on the payoff.

Rules

  • Each candidate is 15-60s, snapped to sentence boundaries.
  • Snap the END to the payoff's last WORD, not to the Whisper segment end.

Segments are pause-based acoustic units — one segment can contain the payoff end AND the first words of the next sentence, and snapping to the segment end drags that fragment into the clip. Use transcript.json's word-level timestamps: end = last payoff word's end + ~0.2s (inside the pause before the next word).

  • Start within the first ~2s of the actual hook line.
  • Rank by hook strength; give a one-line reason and a draft caption.

Return (final message = JSON only)

{
  "role": "hook-finder",
  "section": {"start": 0, "end": 600},
  "hooks": [
    {"start": 132.4, "end": 168.0, "strength": 0.9, "reason": "counterintuitive payoff", "caption": "..."},
    {"start": 410.2, "end": 452.6, "strength": 0.7, "reason": "vivid concrete example", "caption": "..."}
  ]
}