agentleFS
lorena-bordonaba-paulorena-bordonaba-pau/youtube-agent/.claude/skills/likeness-preservation/SKILL.md

likeness-preservation

Skill2 starsChanged 49 days agoMITView on GitHub

Preserve the creator's exact facial identity when generating or editing images. Triggered whenever an image is meant to include a real person.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If a face was not explicitly asked for, do not put one in: When in doubt: objects, text, graphics or abstract.

    line 14
  2. Distinctive:

    Do not proceed without a confirmed reference: If there is none, say so and ask for a photo.

    line 32
  3. Distinctive:

    If it is not recognisably the same person, redo it with a better reference.

    line 55
  4. Distinctive:

    Do not deliver it saying "it came out close".

    line 56
  5. Very specific:

    When generate_image.py detects a reference with the likeness role, it prepends the exact-identity clause and the prohibition on cropping at …

    line 60
  6. Distinctive:

    If the face is not recognisably the creator: say so plainly and ask for a better reference instead of delivering it.

    line 74

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Facial identity preservation

WHEN

Any image that should include the creator, or any real person. Load it before generating, not after seeing the face come out wrong: a lost identity is not recovered by iterating the prompt.

If a face was not explicitly asked for, do not put one in. When in doubt: objects, text, graphics or abstract. An invented person in the thumbnail of a channel with a recognisable face breaks the brand.

EXECUTION ORDER

  1. Get a sharp facial reference. In order of quality:
    • A photo the creator provides in this conversation.
    • An image generated earlier in this conversation where the face came out well.
    • One of their own thumbnails with the face clear and large:
      python3 tools/yt_recent_videos.py --limit 15 --stats
      python3 tools/yt_thumbnails.py --video ID
      And open it to confirm the face is visible. A blurred, profile or obscured face guarantees a bad result.
  2. Do not proceed without a confirmed reference. If there is none, say so and ask for a photo. Generating "something similar" is worse than not generating.
  3. Label it with the likeness role
    python3 tools/generate_image.py --prompt "..." --type thumbnail \
      --ref path_to_the_face.jpg:likeness
    The role does two things: it places the reference first and it activates the exact-identity clause in the prompt. Without the role, neither happens.
  4. When refining, the image to edit goes in --image and the face in --ref face.jpg:likeness. Never the other way round.
  5. Verify by opening the result. The criterion:

    AcceptableNot acceptable
    Different pose, different angleDifferent facial features
    Different expressionSomeone who "looks like them"
    Different light, different clothesA clearly generated face

    If it is not recognisably the same person, redo it with a better reference. Do not deliver it saying "it came out close".

WHAT THE TOOL ALREADY DOES

When generate_image.py detects a reference with the likeness role, it prepends the exact-identity clause and the prohibition on cropping at the neck. There is no need to repeat it in --prompt; verify it with --dry-run.

SOURCES

source: generated. A well-preserved face is still an artefact, not a photograph: if it is used publicly, say so.

OUTPUT

  1. The generated image, after opening it and checking the identity against the table above.
  2. Which reference was used as likeness.
  3. If the face is not recognisably the creator: say so plainly and ask for a better reference instead of delivering it.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.