agentleFS
jammyfujammyfu/open-game-skills/skills/assets/asset-pack-intake/SKILL.md

asset-pack-intake

Skill12 starsChanged 23 days agoMITView on GitHub

Use when a project receives a pre-built genre asset pack (sprite atlases, tilesets, level maps, feel/animation/audio data) described by a pack manifest and must validate it, map every folder to the owning skill, and plan the engine build without editing the art.

The file

Files
SKILL.md

Asset pack intake

Own accepting an external asset pack as input: validate its manifest and files, map each folder to the skill that owns it, and hand the build a phase plan. Do not generate, repaint, repack or edit the art, and do not become a second owner of atlas packing, animation playback, level design or feel tuning.

Modes

ModeOutput
validateintegrity and consistency report with pass / fail / blocked / not-run per check
planphase plan for the chosen engine and art style, at most three specialised skills plus one engine per phase
import-checkrecord of the actual engine import and what it loaded, kept separate from validate

Procedure

  1. Resolve engine, version, target resolution and one art style from the user or project; ask at most one question if the engine or style is unknown. A pack may ship several styles (2D profiles and an optional 3D kit); a build uses exactly one art root at a time but can switch style by id.
  2. Read the manifest contract. Run validate_pack.py on the pack root. It checks hashes, referenced paths, atlas rectangles against real PNG sizes, animation tables, Tiled map layers and tile ids, feel provenance, rights fields and that no executable code ships in the pack. It never decodes pixels, runs an engine or judges art. It also checks the resolution declaration (every 2D style's virtual resolution times its integer scale equals the declared targetResolution; no fractional scale) and the styles.json style map (every asset key resolves to an existing file in every style, atlas animation names exist, GLB nodes and clips exist).
  3. Map folders with the manifest's skills_map; if absent use the table in the contract. sprite-catalog reads the asset manifest, sprite-atlas owns atlas and tileset import, pixel-animation owns tick-based playback, platform-jump / racing-feel / action-feel own feel tables, level-design and level-blockout own levels, art-bible owns palette lock, audio-feel owns cue mapping.
  4. Respect the pack's declared rules. If normalization says characters are re-proportioned to one canonical body per genre, character swaps are cosmetic: gameplay must not read a character's height, build or motion style, and hitboxes come from the data files, never from sprite alpha. Feel is per genre, not per character.
  5. Plan the display and style switching with the engine adapter: the plan output carries the profile's virtual resolution, integer scale and engine settings. All styles of a pack share the same gameplay data (levels and tile ids, tile-unit hitboxes, animation names, frame counts and ticks, feel, entity ids, cue names), so switching style means changing one style id and resolving assets through styles.json; never hard-code art paths and never branch gameplay on the style.
  6. Keep statuses honest. Rows stay draft until imported in the target engine; a packed atlas is not an import, an import is not a playthrough, and a debug-assisted run is not a natural clear. A generated beatability proof is tool-level evidence under the stated assumptions only.
  7. Emit the plan: the validated style and files, unresolved gaps, deferred phases, and the engine adapter's import notes (godot, phaser, unity, custom). Hand other engines the same data through the custom adapter.

Output

Intake report (checks, evidence, assumptions), skill mapping, chosen style root, display and style-switch guidance, phase plan, DEFER list, and a record of build/commit, configuration, scenario, driver and result for every check that ran.

Accept

A pack with a missing file, a changed hash, an atlas rectangle outside its PNG, a map tile id past the tileset, an animation without frames, feel numbers without provenance or a shipped script reports fail and blocks the plan. A pack that passes validate still reports engine import, playthrough and audio perception as not-run until evidence exists. Re-running on the same bytes gives the same report.

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.