More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Edit Timeline Studio: Full Video Editing

    1

    Edit video through an editable timeline project instead of opaque exports: inspect first, plan cuts on paper, lock narration before picture, verify by decoding the render.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Edit Timeline Studio: Full Video Editing

    Edit Timeline Studio: Full Video Editing

    Example session with this skill installed

    I recorded a 40-minute podcast episode on video. Make me a 5-minute highlight reel with captions — the funniest and sharpest moments — that I can post.

    • Read your context and instructions
    • Compiled the edit timeline studio

    HIGHLIGHT EDIT — run through the skill's workflow.

    INSPECT: Source probed, not assumed: 40:12, 1920x1080, 30 fps, two speakers on one stereo track, h264. Brief preserved verbatim; one question asked — aspect ratio (16:9 confirmed, same as source). Deliverable defined as project file plus render, not the render alone.

    PLAN BEFORE CUTTING: A decision record covers the episode in segments. KEEP: the cold-open story (02:10–04:05, high confidence — it sets up the running joke), the argument about AI tasting notes (14:30–17:45, shorten by trimming the middle digression), the closing lightning round (33:00–38:20, keep as-is). REMOVE: the 9-minute equipment tangent (confidence high — no payoff, and the brief says funniest and sharpest). One medium-confidence keep flagged in the report rather than silently decided. Planned runtime is approximate on purpose: captions will follow the actual speech.

    APPLY: Project version saved before the first batch. Picture assembled in one transaction per the decision record; captions generated from the kept segments' speech and checked 1:1 — every visible caption has audible speech for its full interval, no orphan captions over trimmed pauses.

    VERIFY BY DECODING: The render decodes as 1080p30, 5:12, h264 with a real stereo audio track. Boundary pass across all six joins: clock advances monotonically, no repeated clip tails, no black frames. Sampled frames show captions legible and correctly timed at speaker changes. Delivered: the editable project (reorder or restore any segment in seconds) plus the verified render, with the medium-confidence segment named in the report.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Most agent video editing goes wrong the same way: cut first, ask questions later, deliver a render nobody can adjust, and discover the narration doesn't fit the picture after the export. This skill runs editing the other way — the editable project file is the source of truth, and the rendered video is a consequence of it: inspectable, diffable, undoable, re-renderable. The workflow starts with inspection: every asset probed for duration, dimensions, frame rate, audio, and codec, the brief preserved verbatim, and only the genuinely undiscoverable choices (target length, aspect ratio) asked about. Underspecified requests get conservative, reversible decisions with the defaults reported — and a "just open it" request is treated as a handoff, not a license to invent an edit. Execution follows a determinism order — local project-file editing with a structural validation and a semantic dry-run diff before applying, host tools second, browser UI only when nothing else can express the operation, with the limitation stated. Planning happens at the right fidelity: automatic edits get a written decision record (keep, remove, shorten, reorder — with reason and confidence for every segment) before a single cut, and narrated edits follow the audio spine: narration written as breath-group segments, synthesized to separate files, locked first, with caption and picture timing derived from measured speech — never speech crushed to fit a planned runtime. Application is transactional and versioned. Verification decodes the actual render — boundaries, caption-to-speech resolution, audio behavior, container facts — under a freshness rule that forbids presenting a pre-fix render as evidence of the fix. A worked example turns a 22-minute tutorial into a 5:47 narrated highlight, and a delivery scorecard gates the handoff.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Video Skills.

    • It edits existing footage; it is not a footage-generation skill, and generation enters only as an explicitly requested step inside an edit.
    • Verification by decoding proves the render matches the project and the brief; it cannot judge taste — the decision record exists so a human can overrule the editor's calls before or after delivery.
    • Real-time and live production are out of scope: the workflow assumes recorded media and a re-renderable project.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 24 days ago

    Permissions

    No special permissions declared or detected

    Listed24 days ago
    Updated21 days ago

    What's inside

    Frequently Asked Questions

    Because the first version is rarely the last word. With the project as the deliverable, "swap segment three," "restore that joke," or "make it 4 minutes" are small transactions and a re-render — without it, every revision starts from the raw footage again. You get both files every time.

    It does — but it writes the decision record first: every segment marked keep, remove, shorten, or reorder with a reason and a confidence level, protecting the content your format can't lose (the setup a highlight depends on, the punchline a tutorial builds to). You see the reasoning before the cut, and medium-confidence calls are flagged to you instead of buried.

    It reverses the order. Narration is written as short breath-group segments, synthesized to separate files, and locked; caption and picture timing are then derived from the measured speech. Speech is never sped up or stretched to fit a pre-planned runtime — the most common cause of the misfit you're describing.

    New in Image & Video