kling-promptinglisted

# Kling Prompting Skill ## When to load this skill Load this skill before invoking any `media_kling_*` MCP tool or before dispatching to the `kling-director` subagent. The default video-router uses this skill's tier-selection guidance to decide which Kling model to use. ## 5-part prompt spine (use for t2v, i2v, motion-brush, lip-sync, extend) Every Kling prompt should answer five questions in order: 1. **Camera** — lens, height, angle, motion verb. Examples: "85mm portrait, eye-level, slow dolly-in" / "wide 24mm, low-angle, static" / "handheld shoulder-cam, tracking right". 2. **Scene** — location, time-of-day, weather, light direction. Examples: "rooftop bar at golden hour, west-facing sun, scattered clouds" / "abandoned warehouse, blue hour, lone overhead practical light". 3. **Action** — subject + verb + object. Singular, present-tense, no hedge words. Examples: "woman in red coat lights a cigarette" / "crow lands on the rusted hood". 4. **Vibe** — adjectives for mood + film stock + grade. Examples: "melancholic, Kodak Portra 400, desaturated cyan-orange" / "tense, anamorphic flares, high-contrast green grade". 5. **Time** — duration + pace + rhythm. Examples: "5s, deliberate, single beat" / "10s, two-beat: setup then reveal". Concatenate with commas. Do NOT use bullet lists in the prompt itself — Kling parses better with a single descriptive sentence stream of 30-80 tokens. ### Worked example Spine: - Camera: "50mm, chest-height, slow truck-left" - Scene: "neon-li