skillZs
★ LIVE SKILL TAGS ★
>>> LIVE SKILLS INDEX <<<
* OPEN SOURCE *
NO LOGIN, NO TRACKING
※ REAL INSTALL DATA ※
← back to all skills
mojomast/open-higgsfield-skills68 installs

continuity-check

Sweep adjacent shots and takes for continuity breaks (cast, wardrobe, props, grade, lighting, screen direction, timing) at plan level and take level, and report findings with proposed fixes - never edits. Use before locking a cut, after a take is replaced, or when asked "do these cut together?".

How do I install this agent skill?

npx skills add https://github.com/mojomast/open-higgsfield-skills --skill continuity-check
view source ↗

Is this agent skill safe to install?

  • Gen Agent Trust Hubpass

    The skill is designed for checking continuity in video production and appears legitimate. It utilizes standard tools like ffmpeg for frame extraction and includes robust safety rules, such as prohibiting direct file modifications and requiring cost estimates for generation requests. The primary security consideration is a low-level risk of indirect prompt injection, as the skill processes external production data (e.g., story text and notes) that could potentially contain adversarial instructions, though the skill includes specific directives to mitigate this behavior.

  • Socketpass

    No alerts

  • Snykpass

    Risk: LOW · No issues

What does this agent skill do?

continuity-check

This follows ZeroField's continuity-check (MIT). The full method is in references/zerofield-continuity-check.md.

Procedure

  1. List every join (shot N to shot N+1) and mark the ones touched by recent changes.
  2. Plan level: for each join, compare the plan entries. Cast and wardrobe ids, props, location and time of day, screen direction (the 180° rule), the camera axis, and the grade or LUT should agree unless the plan marks a deliberate change.
  3. Take level: for each join, compare the last frame of N with the first frame of N+1. Extract them with ffmpeg (-sseof -0.05 -frames:v 1 / -frames:v 1). Use a VLM: still-qa/scripts/judge.py with --ref set to the other frame, the prompt "same character, wardrobe, props, lighting, direction?", and the judge scores from still-qa.
  4. Classify each finding as break (must fix), drift (should fix) or note, with evidence: the frame paths and the plan fields involved. If you have no evidence, report nothing.
  5. Propose the cheapest fix: edit the prompt, re-seed, image-edit the keyframe, chain from the last frame, or re-grade. A proposal names its target skill and parameters.
  6. Report breaks first, then drift. The user or the orchestrator decides what to apply.

Hard rule

This skill never writes to the plan, the files or the renders. It returns findings and proposals only.

Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.

<a href="https://skillzs.dev/skills/mojomast/open-higgsfield-skills/continuity-check">View continuity-check on skillZs</a>