Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add YansIlinta/video-agent-runtime --skill video-editinggit clone --depth 1 https://github.com/YansIlinta/video-agent-runtimeWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/yansilinta/video-agent-runtime/video-editing)<a href="https://agentmods.dev/skills/yansilinta/video-agent-runtime/video-editing"><img src="https://agentmods.dev/badge/skills/yansilinta/video-agent-runtime/video-editing.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00060 | $0.00911 |
| Opus 5 | $0.00030 | $0.00456 |
| Sonnet 5 | $0.00012 | $0.00182 |
| Haiku 4.5 | $0.00006 | $0.00091 |
Grade A, and why
video-editing scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 62 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Video editing runtime
Use the runtime as the project authority. Do not edit its JSON files, construct FFmpeg commands, or bypass validation.
Start with state and evidence
- Call
project_status. - Call
system_statusbefore choosing a real provider path; unavailable optional alignment or TTS is not permission to pretend it ran. - Read the active transcript,
transcript_quality, and timeline before proposing changes. Prefertranscript_search,transcript_inspect_range, andtimeline_inspect_rangeover re-reading large artifacts. - Call
visual_inspect_rangeonly when the decision genuinely needs visual evidence. Do not inspect every frame or send the whole video to a vision model by default. - Surface low-confidence words, failed alignment, speaker overlap, or other transcript-quality warnings when they affect an editing decision.
Approval-gated edit loop
For a new edit direction:
- Understand the requested outcome and constraints.
- Call
strategy_proposeand explain its structure, target duration, selection policy, pace, captions, and rationale. - Wait for explicit user approval. Then call
strategy_approve. - Call
edit_plan_create,edit_plan_validate, andedit_plan_diffin that order. - If validation fails, fix the structured plan; never work around it in the renderer.
- Show the meaningful diff and apply only after it is consistent with the approved strategy.
- Call
edit_plan_apply, thenpreview_render. - Review the self-check result and ask the user to assess the preview.
- Call
final_approveandexport_videoonly after explicit final approval.
Never treat self-evaluation as user approval.
Feedback and replanning
- Submit the user's original words with
feedback_submit; include category and range when known. - For a local complaint such as “20–35 seconds is slow,” prefer the smallest plan change affecting that range. Preserve unrelated clips.
- Call
workflow_diagnosebefore repeated retries. - Use
PATCHfor a localized execution problem. UseREPLANwhen feedback indicates the EditingStrategy is wrong. - For
PATCH, calledit_patch_plan,edit_patch_validate, andedit_patch_diff; reject out-of-scope/global mutations beforeedit_patch_apply. - When diagnosis returns
REPLAN, callworkflow_replan, present the replacement strategy, and wait for a fresh approval. - Repeated “flat,” “not interesting,” or “weak opening” feedback is evidence of a possible story-structure mismatch, not permission to randomly move cuts again.
- Explain what changed, why it changed, and which feedback caused it. Use version comparison when that is clearer.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 7d ago First seen · 62 lines · 60 tokens per session scan A 1c6e9af71b24
video-editing is a skill published in the GitHub repository YansIlinta/video-agent-runtime (1 stars, last pushed 13d ago), licensed MIT. It adds 60 tokens to every session and 911 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
vidgrid
Give an AI agent eyes for video. Turn a local video file (MP4, MOV, WebM) into a numbered grid of frames plus a transcript that a vision model can read — so you can summarize a talk, find the exact moment something happens, read on-screen text or UI, or rank clips. Use whenever you need to watch, analyze, or…
claude-real-video
Watch a video for the user. Use when the user shares a video URL (YouTube etc.) or local video file and wants it summarized, analyzed, or discussed — Claude can't ingest video directly, so this skill extracts scene-aware keyframes + transcript first, then reads those.
edit-video
Edit video with FableCut — assemble a cut, add titles and captions, grade, add transitions, keyframe animation, speed ramps, and export. Use whenever the user wants to cut, trim, caption, subtitle, grade, or otherwise edit a video, or asks to open/start the video editor.
remake-reel
Analyze a reference video (a reel, montage, or ad the user likes) into an edit blueprint — shot boundaries, music beats, BPM, energy curve, the drop — and rebuild the same structure with the user's own footage. Use when someone says "make it like this video", "remake this reel with my clips", or asks what makes an…
claude-real-video-for-agents
Install and use crv (claude-real-video) — a tool that lets any AI agent watch videos by extracting scene-aware keyframes, deduplicating them, and transcribing audio. Use when the user shares a video URL or file and wants it analyzed, summarized, or discussed.
authoring-video-templates
Use when creating or editing a video template JSON (the template descriptor), adding or changing sections/filters/maps/variables/transitions/looks/motion/audio/layers, or debugging template validation errors in ffmpeg-video-composer.