Who it is forA creator with a character who appears in many scenes and whose face, clothing, or age could drift between generations.
The problemA passport built from one lucky image looks consistent in one clip but fails across angles, lighting, and co-stars.
The outcomeA passport with a verbatim descriptor, separate tagged state variants, and a registry row set to locked after a 10 out of 10 repeatability pass.
What you need first
- An agent harness that loads skills: Claude Code, Codex, Hermes, or OpenCode
- Access to the image model and video model you plan to use, named during /setup
- A script, treatment, or one-paragraph idea for the film
- Reference images for characters and locations, if you have them
Steps
- Run
/asset-passportand name the character, then answer its questions until the descriptor has no gaps. - Generate the grey-background reference-sheet images from the prompts the skill writes, using your image model.
- Run
/stress-testto build the combat matrix and test prompts for the character. - Check that the registry row flips to locked only after the character passes 10 out of 10 repeatability.
Prompt for your agent
Using the film-studio-skills repo, run /asset-passport and /stress-test for the character <character name>. Expected output: a passport with a full descriptor and a registry row marked locked or draft. Verify that the row is locked only if all test prompts scored 10 out of 10.
Caveat. The stress test uses your own image model outputs, so the lock is only as reliable as the test images you actually generate and score.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for machina-exm/film-studio-skills and "Lock a recurring character before filming their shots"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer