Skip to content
Back to skills

Create Kling Scene

ASecurity

Compose a Kling video request from a pydantic-validated scene table and per-character reference images, with no bespoking. Use when a user says create a kling scene, make a kling video from a scene table, compile a scene script to kling, or wants the scene-writing -> kling pipeline run as one gated command instead of hand-built requests.

  • 6 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added September 11, 2026
content-marketingpythonbash

Security analysis

A100/100

Pro scans all 6 files and shows the line behind each finding

Scanned September 11, 2026

npx -y skills add grahama1970/agent-skills --skill create-kling-scene --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Create Kling Scene?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Create Kling Scene
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/grahama1970-create-kling-scene/badge)](https://www.skillsdirectory.com/skills/grahama1970-create-kling-scene)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: create-kling-scene
description: >
  Compose a Kling video request from a pydantic-validated scene table and
  per-character reference images, with no bespoking. Use when a user says
  create a kling scene, make a kling video from a scene table, compile a
  scene script to kling, or wants the scene-writing -> kling pipeline run
  as one gated command instead of hand-built requests.
triggers:
  - create a kling scene
  - make a kling video from this scene
  - compile scene table to kling
  - kling scene pipeline
provides:
  - kling-scene-composition
composes:
  - interview
  - best-practices-scene-script-writing
  - best-practices-kling-video
  - triage-error
  - agentic-evals
complies:
  - best-practices-skills
  - best-practices-python
  - best-practices-kling-video
  - best-practices-scene-script-writing
runtime_self_improvement: basic
---

# create-kling-scene

One gated command from scene script to submittable Kling packet. Creating a
Kling video no longer requires bespoke request assembly — it requires two
specific inputs and everything else is deterministic:

```text
INPUT 1: scene table   (scene_script.scene_table.v1 — must pass the
                        scene-script-writing pydantic gate)
INPUT 2: references    (one single-subject crop per character row —
                        validated as real decodable PNG/JPEG, magic bytes,
                        >=1KB, not just an existing path)
INPUT 3: voice wavs    (optional --voice name:purpose=/path.wav — validated as real
                        RIFF/WAVE within Kling bounds: 2-60s lipsync,
                        5-30s voice_clone)
   |
   v
./run.sh build --scene scene.json --refs embry=/path/embry.png --out-dir /path/run
   |
   1. scene_table_gate    scene-script-writing pydantic validate (fail-closed)
   2. reference_check     every character row has exactly one existing ref
   3. instructions_gate   create_kling_scene.instructions.v1 — typed FINAL
                          instructions (per-slot char budgets, @ElementN
                          binding coverage, character<->ref<->voice cross-
                          checks, required_fact_ids coverage) validated BEFORE any Kling-format
                          conversion; emitted as kling_instructions.json
   4. compile             MECHANICAL instructions -> kling_video.request.v1
                          (no decisions below the validated layer)
   5. kling_packet_gate   kling-video pydantic validate (endpoint/element/
                          budget/URL rules)
   |
   v
receipt.json (create_kling_scene.receipt.v1) + kling_request.json
+ next_command for the PAID submit (never auto-submitted)
```

Every stage failure carries pydantic `errors[]` AND a triage-error
classification (`{code, cause, next_command}`); ambiguous signals self-heal by
minting a provisional catalog code. When triage cannot fix it (unrecoverable,
freshly minted unclassified code, or classifier unavailable), the receipt gains
`needs_attention` with an $interview handoff: `interview_questions.json` is
written beside the receipt and `resume_hint` runs the interview. The receipt is
typed (`create_kling_scene.receipt.v1`, pydantic extra=forbid, producer-side
seam_validation stamp), never prose.

## Fact coverage contract

All environment header fields and every element row in the scene table become `required_fact_ids` by default.
`kling_instructions.json` must list the same IDs in `covered_fact_ids`; otherwise `instructions_gate` BLOCKS before any provider packet is written. Lower-priority mood/style cuts belong in `dropped_desired_fact_ids`, never by silently losing required rows during prompt condensation.

## What this skill refuses to do

- Accept prose instead of a scene table (no table -> `scene_table_gate` BLOCKED).
- Accept a character without a reference image (`reference_check` BLOCKED).
- Describe character identity in prompt prose (identity travels in `elements[]`).
- Submit. `build` stops at a validated packet; the paid call is the human/agent
  running the emitted `next_command` (kling-video `submit`), per Rule 9
  no-silent-retry.

## Commands

```bash
./run.sh build --scene scene.json --refs embry=/abs/embry.png horus=/abs/horus.png --voice embry:lipsync=/abs/embry.wav --out-dir /mnt/storage12tb/skills/create-kling-scene/outputs/<run>
./sanity.sh   # offline positive + negative fixtures
```

## Ownership

Scene writing rules: best-practices-scene-script-writing. Kling mechanics,
budgets, and the paid submit: best-practices-kling-video. Failure taxonomy:
triage-error. This skill owns only the composition order and the receipt.

Files in this skill

  • SKILL.md4.5 KB
  • fixtures/agentic_eval.json5.7 KB
  • references/webgpt_review_2026-09-08.md26.1 KB
  • run.sh312 B
  • sanity.sh6 KB
  • scripts/create_kling_scene.py18.9 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.