- Home
- Skills
- Image & Video
- Stage Assemble: Video Plan to Finished File
More screenshots
Works with the AI Agents you already use
Stage Assemble: Video Plan to Finished File
Assemble an approved multi-source video plan into a finished file: checkpointed segments, layered assembly, narration mixed exactly once, and a QA gate before delivery.
$5
Stage Assemble: Video Plan to Finished File
Example session with this skill installed
Assemble this approved plan: a 60-second vertical launch video, 5 segments (2 footage edits, 1 generated clip, 1 composed title card, 1 provided outro), narration across 4 windows, captions supplied, canvas 1080×1920 at 30 fps. One generated clip is billable.
- Read your context and instructions
- Compiled the stage assemble
RUN
- Pre-flight: plan validated, caption data final, narration windows do not overlap, all sources present, billable generation count (1) confirmed before anything produces. Journal opened: 5 segments pending.
- S1 footage edit — probed 14.8 s vs 15.0 planned, on-canvas: verified, warn absorbed by the following card.
- S2 generated clip — produced once, 6.0 s, silent as required: verified.
- S3 composed title card — arrived with a voiceover baked into its render: REJECTED at intake under the exactly-once narration rule. Re-rendered silent, journal updated, no other segment touched.
- S4 footage edit — 12.1 s: verified. S5 provided outro: verified.
- Assembly, layers in order: primary track conformed to the declared 30 fps canvas, the full-screen card placed in the primary track (not as an overlay), narration mixed exactly once across the 4 windows, one music bed ducked under the voice, captions burned last from caption data, loudness normalized and measured: −14.0 LUFS integrated, −1.0 dBTP true peak.
QA GATE — Technical: pass (60.2 s, on-canvas, audio present). Promise: pass (real motion 78% of runtime vs 65% committed). Visual: pass, 10 frames pulled across the timeline, no garbled text or empty frames. Audio: pass, narration coverage full, no overlaps, transcript matches the plan.
DELIVERED — the finished file, the journal (5 verified, 1 rejection, 1 re-render), and this QA report. The one defective segment cost one re-render — not a rebuild of the other four.
Connects securely to your tools. The creator never sees your data.
About this skill
The assembly discipline for multi-source video: turning an approved plan — footage edits, generated clips, composed graphics, supplied pieces — into one finished file without losing anything the plan promised. It is built against the three enemies of assembled video: rework, double audio, and silent drift. A checkpoint journal records every segment's status and output path the moment it verifies, so a crash or a revision at 90% costs 10%, not a rebuild; billable generations get an extra authorization guard. Segments are produced by source in plan order and probed on arrival — duration, resolution, audio presence — with numeric intake scoring that rejects off-spec pieces at the door instead of at final QA. Assembly runs in a fixed six-layer order (primary track, overlays, narration exactly once, one continuous music bed, captions burned last from data, loudness normalization with measured numbers) because the order itself is the correctness argument: most assembled-video defects, including the signature two-voices-saying-the-same-line failure, are layer-order violations. A continuity section covers how mixed sources read as one video, a worked run traces a 90-second explainer through one rejection and one send-back, and no draft ships without a written four-section QA report — technical probe, promise check, visual spot-check, audio check — with bounded send-backs to the single segment at fault.
- It assembles an existing approved plan; it does not write the plan or generate footage itself — planning and generation are separate stages.
- The visual spot-check requires the ability to inspect frames; where that is unavailable the skill marks the check unverified rather than guessing.
- Loudness targets follow common delivery practice; a specific platform or broadcaster spec you must hit should be stated in the plan.
How to use
- 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
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
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