agentleFS
MicrosoftKnown publishermicrosoft/amplifier-acp-minecraft/skills/minecraft-agent-skins/SKILL.md

minecraft-agent-skins

Skill1 starsChanged yesterdayMITView on GitHub

Create, refine, validate, and install recognizable Minecraft agent skins from portraits or existing textures. Use for Minecraft player skin artwork and texture mapping, not general portraits or game behavior.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    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:

    Read artwork-and-layout.md for a generation brief, UV conventions and failure examples.

    line 34
  2. Distinctive:

    Do not sharpen the entire atlas indiscriminately.

    line 43
  3. Distinctive:

    Keep personal references, generated artwork, prompts and installation receipts in the owner's private asset directory.

    line 47
  4. Very specific:

    If inspected opaque artwork has only near-opaque export noise (for example alpha 252–254 throughout sampled base faces), the private map …

    line 73
  5. Distinctive:

    Read installation.md for account versus server installation, signing, refresh and evidence.

    line 88
  6. Distinctive:

    When a real failure reveals a repeatable technique, update the relevant guide or helper here with the fix and its limits.

    line 104
  7. Distinctive:

    Keep generic image-generation instructions in the host's image tooling.

    line 108

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Minecraft agent skins

Turn an agent's approved identity into a readable player skin. Use the host's available image-generation/editing capability for artwork; this skill supplies Minecraft-specific constraints and verification. It does not require a particular model, provider, ACP engine, or central provisioning service.

Choose the target and preserve identity

Inspect the portrait and any existing skin before editing. Keep the identifying palette, face shape and one or two recognizable accessories. Ask only if the identity or target is missing. A skin refresh changes appearance, not the agent's identity, session, engine, or construction authorization.

For a Java-compatible shared-world skin, start with a 64×64 classic (four-pixel arms) PNG, unless the existing character or user calls for slim. The included utilities intentionally support classic only. Do not silently label a slim texture classic or assume a Bedrock Character Creator's extra geometry can become a flat Java skin. A beard, spectacles or clipboard must be suggested within the available head/body pixels and optional outer layer.

Design for the pixels players see

The face is 8×8 pixels, not a miniature high-resolution portrait. Keep eyes distinct from skin, hair and frames. Prefer deliberate color blocks and dark/light separation to gradients and tiny decorative marks. Check accessories at native size; simplify them when they aren't recognizable.

Read artwork-and-layout.md for a generation brief, UV conventions and failure examples. A generated image is artwork, not proof of a valid atlas. Even a prompt listing exact UV coordinates can produce uneven panels or the wrong dimensions.

Large artwork can look attractive but produce muddy eyes when averaged down. Nearest-neighbor sampling preserves existing color boundaries; it cannot recover details absent from the target grid and can lose narrow features. Review the resulting face instead of assuming a resampling algorithm solved readability. Do not sharpen the entire atlas indiscriminately.

Pack and validate locally

Keep personal references, generated artwork, prompts and installation receipts in the owner's private asset directory. Publish reusable improvements, generic fixtures and helpers in this repository.

From this repository's root after npm ci --ignore-scripts:

node skills/minecraft-agent-skins/scripts/skin-atlas.cjs validate /absolute/skin.png

For an approved, base-layer-only generated atlas with noncanonical panels, make a private JSON map with width, height, and parts. Use the part names and six-face order in classic-uv.json; each source rectangle is [x,y,width,height]. Measure the actual source, excluding gutters and labels. Then:

node skills/minecraft-agent-skins/scripts/skin-atlas.cjs pack /absolute/source.png /absolute/layout.json /absolute/new-skin.png

The packer samples selected faces without blending, creates an exact classic base atlas, rejects transparent sampled base pixels and refuses overwrite. It does not infer panels, fix artwork, preserve an existing outer layer, mirror limbs, or prove correct face orientation. For an existing skin with overlays, edit that texture while preserving its layout instead of using this base packer.

If inspected opaque artwork has only near-opaque export noise (for example alpha 252–254 throughout sampled base faces), the private map can explicitly set minimumBaseAlpha from 250 to 255. Default 255 remains strict. The packer rejects samples below that threshold, changes only accepted base alpha to 255, preserves RGB, and reports the normalized pixel count. This is not a repair for holes, intentional transparency, incorrect crops or overlays. See the source alpha first.

Validate dimensions, opaque base faces and partial-alpha diagnostics. Inspect the 8×8 face at native size and enlarged with nearest-neighbor, then view the mapped character from front, back and both sides. Verify top/bottom, limb seams, hands, boots and whether overlays hide the eyes. Numerical validation does not establish resemblance or visual quality. Compare revisions under similar lighting.

Install only when requested and verify

Read installation.md for account versus server installation, signing, refresh and evidence. Inspect the actual server mechanism; do not assume a plugin, world path, version or UUID.

Creation alone need not install anything. When installation is authorized, preserve current player/task/session state. Prefer a supported skin refresh. Coordinate an idle boundary if a reconnect/restart is necessary; do not restart the world or other agents for a cosmetic update.

Deliver the final PNG, a useful mapped preview, selected model, and what was actually checked. Separate file validation, server texture delivery and client appearance. Health checks alone prove neither delivery nor appearance. Record pending visual verification honestly.

Improve from evidence

When a real failure reveals a repeatable technique, update the relevant guide or helper here with the fix and its limits. Add regressions for mapping/validation defects; keep personal before/after evidence private. Do not hard-code one person's face, host, player name or credentials. Do not call an untested technique proven. Keep generic image-generation instructions in the host's image tooling; keep Minecraft UV, pixel readability and skin delivery in this adapter skill.

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.