Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →openakita/footage-gate
Review source media, cut silence, auto color grade, and run cut-boundary QC through Footage Gate.
Passed 20 of 21 test cases with the skill, and 3 of 21 without it.
and 37 more in the full report
18 fixed3 passed either way1 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 17 test cases the skill fixed. Same model, same prompt, with and without the skill.
Write a standalone command line invocation to apply the subtle color grading equalizer filter directly to `raw_take.mp4` via terminal.
| Checked | Without skill | With skill |
|---|---|---|
| Uses ffmpeg CLI tool with eq video filter | Passed | Passed |
| Contains exact string contrast=1.05:gamma=1.02:saturation=1.04 | Missed | Passed |
The instructions your model reads when the skill loads.
Operator manual for the
footage-gateplugin (v1.0.0). Use this when you are running the plugin in production or coaching another agent through the workflow. For source-code architecture seeREADME.md§2.
A four-mode local post-production quality gate. Each mode is a fully
deterministic FFmpeg pipeline plus a thin Python wrapper that surfaces
risks, fixes, or both. There is no LLM involvement in the default path —
the only optional cloud call is a DashScope Paraformer transcription
under source_review.
The plugin owns its own SQLite (tasks + config + assets_bus reserved
for v2.0), its own FFmpeg installer panel (1:1 reused from
seedance-video), and its own React UI (single-bundle, ≤2800 lines,
8 hard contracts aligned with tongyi-image).
| Situation | Mode | Why |
|---|---|---|
| You just imported the day's raw footage and want a triage report. | source_review | Catches resolution / mono-audio / too-short clips before you start editing. |
| A long take has obvious dead air you want trimmed. | silence_cut | Pure-NumPy RMS detection, no aubio install pain. |
| A finished cut feels flat / too dim. | auto_color | Sub-second sampling → ±8 % clamped grade. HDR sources auto tone-mapped. |
| You exported the master and want to verify QC before delivery. | cut_qc | 4 checks vs your EDL, with optional auto-remux loop (≤3). |
Combine them as the recommended workflow:
source_review → silence_cut → auto_color → cut_qc. Each step is
its own task and its output is the next step's input.
| Mode | Accepted inputs | Output artifacts |
|---|---|---|
source_review | mp4 / mov / wav / mp3 / jpg / png | report.json (risks, ffprobe metadata, optional transcript), 4 representative frames |
silence_cut | mp4 / mov / wav / mp3 | output.mp4 (or .wav), report.json (intervals removed) |
auto_color | mp4 / mov | graded.mp4, report.json (chosen eq values, HDR flag) |
cut_qc | mp4 / mov + EDL JSON | report.json (issue list per check), optional remux_N.mp4 for each attempt |
test cases passed, vs 3 / 21 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.