premiere-pro-mcp-editing

PPMCP agent skill — automatic systems over tool count

Quality bar

A long chain of atomic calls is not a better cut than one playbook that already knows the professional defaults. Optimize for cut quality, not call count:

Capability	Use
Automatic edit	edit_auto + playbooks — one call, complete structure
Editable titles	text engine: UXP → hybrid → CEP → PNG fallback
Delivery polish	quality_pass / edit_delivery — grade + transitions + fades
QA gate	edit_verify before export
Failures	recovery hint, then continue the plan
Token cost	playbooks + compact: true + the standard tool profile

Never sacrifice quality for “one more atomic tool.” Prefer automatic playbooks that already include grade + transitions + fades (+ normalize).

Session flow

edit_bootstrap
  → edit_auto (intent)  OR  edit_playbook_run (named pack)
  → edit_verify
  → (strong model) atomic polish
  → export / save

Prompt templates (copy-paste)

Full system + user prompts: docs/PROMPTS.md

Track	Tools	Use when
Weak	only edit_bootstrap + edit_auto + edit_help	small models, low token
Strong / premium	playbooks first → atomics + vision QA	client review, trim, captions

Weak model (cheap, still good)

Only:

1. edit_bootstrap
2. edit_auto with intent + media args
3. edit_help if stuck

{
  "intent": "youtube episode",
  "args": {
    "paths": ["C:\\\\media\\\\clip.mp4"],
    "title": "Episode 1",
    "lowerThird": "Alex Host",
    "musicPath": "C:\\\\music\\\\bed.mp3",
    "look": "warm"
  }
}

{
  "intent": "tiktok",
  "args": {
    "paths": ["C:\\\\media\\\\clip.mp4"],
    "title": "WAIT FOR IT",
    "sfxHits": [{ "path": "C:\\\\sfx\\\\whoosh.wav", "atSeconds": 1.2 }],
    "look": "warm"
  }
}

Strong model (premium)

1. Same automatic pack for structure + polish (edit_auto / edit_playbook_run)
2. Then atomics: clip_trim, workflow_ken_burns, color_set_param, audio_set_gain, silence/captions
3. Vision: sequence_screenshot / edit_verify
4. Export with explicit path/preset when user asks