Local video editor for coding agents: 42 FFmpeg tools with a machine-readable contract and an MCP server, for Claude Code, Cursor and Codex. Cut, captions, reframe, loudness, platform exports. No cloud, no API keys.
Trigger tests: should / should-not
prompts.json holds 26 requests the skill must trigger on (English, Japanese, Chinese, Korean, Spanish, Arabic, Thai, Hindi, Hebrew, Russian, Vietnamese and Turkish, some without a file extension or the word "edit") and 19 near-miss requests it must not (still images, installing ffmpeg — in English, Japanese, French and Greek, downloading from YouTube, writing a script in Japanese, Portuguese and Italian, an editor shortcut, thumbnail design, CSV to JSON, a codec explainer, AI video generation in English, German and Indonesian, meeting transcription, an emoji sticker pack as PNGs — image work, not video, chapter titles for a blog outline with no media, and designing a podcast cover, which is image design rather than an edit). 1.16 added three that must trigger: a podcast episode to post on YouTube, a Japanese request for chapter marks with no idea where to cut, and a Spanish request for switchable English and Spanish subtitle tracks. 1.17 added five more that must trigger, one per feature and none naming a tool or a flag: captions coming out in three lines with half the sentence late (en), cutting to the BGM's beat (ja), stripping the ums and uhs out of a podcast recording (en), 200 videos in a folder taking an eternity (es), and "I only changed the export preset, do I have to re-render the whole edit?" (en) — so the set was 31 should-trigger and 19 near-miss, 50 in all. 1.18/1.19 added five more that must trigger, one per 1.18.0 analysis/multicam feature and again none naming a tool or a flag: is the camera locked off or drifting (en, cropdetect.py's motion signal), keep the breaths but cut the odd long silence (ja, silence.py --speech-aware), where's the motion so a 9:16 crop doesn't cut off the action (en, cropdetect.py --motion-centre), three cameras at once, always cut to whoever's loudest (es, multicam.py --switch energy), and three unsynced phone recordings of the same panel (en, sync.py with more than one extra source) — so the set is now 36 should-trigger and 19 near-miss, 55 in all.
Method: show an independent model a skill catalog made of this skill's description plus four decoy skills that own the neighbouring territory (image-tools, web-research, copywriter, meeting-notes), give it the raw request, and ask which one skill it would load or none. A run is correct when ffmpeg-skill is chosen exactly for the should-trigger prompts.
The non-English prompts are there because the skill description is English: a request in Chinese, Korean, Spanish, Portuguese, French, German, Arabic, Thai, Hindi, Hebrew, Russian, Greek, Vietnamese, Indonesian, Turkish or Italian must route the same way an English one does, and a near-miss in those languages must still miss.
Results: results-<date>.json. 2026-09-04: 22/22 on the then 22-prompt set (incl. two audio-only requests: WAV→MP3 conversion, silence removal on an M4A). The set is now 40 prompts (22 + 7 multilingual: t13 zh, t14 ko, t15 es, t16 ar should-trigger; n11 pt, n12 fr, n13 de should-not; then 9 more: t17 th, t18 hi, t19 he, t20 ru, t21 vi, t22 tr should-trigger; n14 el, n15 id, n16 it should-not; then 2 for 1.15: t23 an emoji caption request should-trigger, n17 an emoji sticker pack as PNGs should-not — the neighbouring territory image-tools owns). 2026-09-14 (it18): 50/50 on the full 50-prompt set — 31/31 should-trigger, including the five new 1.17 prompts, and 19/19 near-miss. 2026-09-14 (it19): 50/50 again on the same 50 prompts at skill 1.17.1 — 31/31 should-trigger and 19/19 near-miss, unchanged. 2026-09-14 (it20): 49/50 at skill 1.17.2 with the same description text — the judge sent t49 (es, "200 videos in a folder take forever, is there a faster way?") to none as "vague, no actionable file"; 31/31 minus that one, 19/19 near-miss. Same description, different answer: judge variance on the one prompt that names no file, not a description change.