agentleFS
NewenergyEricNewenergyEric/art-director-skill/skills/game-art-production/SKILL.md

game-art-production

Skill5 starsChanged 37 days agoMITView on GitHub

Continue game-art production from a director's project pack. Assign and create related characters, environments, props, UI and animation, review them in context, and export optional Compositor scenes without inventing another visual style.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    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. Very specific:

    Read the pack's SKILL.md, style-pack.json, manifest and current task.

    line 10
  2. Distinctive:

    Use the production contract for plan fields and recovery.

    line 36

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Game Art Production

Use the direction established in the project pack to make consistent components, whoever creates them. art-director establishes and revises the shared judgment; this Skill applies it to production and assembly. A received project pack is sufficient to start. If direction is missing, resolve that gap before making a batch; do not inherit Tidekeeper's look by default.

Read the pack's SKILL.md, style-pack.json, manifest and current task. Verify the files and pin its version, fingerprint and manifest hash. Work in the user's language. The agent handles records and tools; designers should not need to fill JSON before creating art.

Continue the work

  1. Establish the player action, intended feeling and actual display context. Split the work into necessary assets, states and assemblies. Richness comes from meaningful relationships and behavior, not a count of decorative objects.
  2. Assign each asset to a person or agent and identify its upstream first article, shared rules, identity references, permitted changes and destination composition. Designate who assembles the result. Record these in the task brief or existing batch record; assignment is not completion. Reuse verified unchanged outputs, and keep missing, rejected or obsolete inputs visible.
  3. Actually attach relevant images to generation or editing tools. Save the prompt, tool, inputs and returned files. Check a representative result before extending its family. The scripts provide deterministic checks and exports; they do not call a model or certify aesthetics.
  4. Compose at the intended scale. Review shape language, edges, material, lighting and type together with character contact, contrast, layering and UI priority. A matching palette alone is insufficient. Typography should belong to the pictured world while gameplay text remains readable. Remove repeated copy that the artwork already communicates.
  5. For movement, choose an appropriate method and inspect actual playback. A short action may use 6–20 distinct drawn/generated poses or deformation frames, a rig, or generated animation when a suitable tool exists. This range is a production option, not a frame-count gate for every effect. Repeating one displaced image does not provide the missing poses. Follow assembly and animation review.
  6. Return recurring direction gaps to a candidate revision of the project pack; keep single-asset fixes in their tasks. Use --allow-candidate for candidate handoffs without turning a technical pass into director approval. Existing user authorization remains valid.

Delivery paths

Standard production: project pack → representative asset → related assets → actual composition and playback → review and delivery. Use the available tools and target engine; Compositor is optional.

Layered editing: export the same scene recipe to .comp after asset production. Read Compositor support. It is a layered editing destination, not a generation service, game engine or animation renderer. Windows can export and inspect the format. Record native acceptance only after opening, inspecting and saving on a supported Mac.

scripts/production.py requires Python 3.10+. check, tasks and verify use the standard library; PNG conversion and Compositor export require Pillow. Outputs must use new directories, preserving members' working files.

python scripts/production.py check PLAN.json --pack RECEIVED_PACK --allow-candidate
python scripts/production.py tasks PLAN.json --pack RECEIVED_PACK --out NEW_TASKS
python scripts/production.py build PLAN.json --pack RECEIVED_PACK --out NEW_DELIVERY --allow-candidate
python scripts/production.py verify NEW_DELIVERY

Use the production contract for plan fields and recovery. Report what was made, the shared direction used, actual observations, file entry points and unresolved work. Keep candidate art, technical validation, self-review, director decisions and native or engine acceptance distinct.

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.