0x0funky/agent-sprite-forge / skills/generate2dmedia
generate2dmedia
VerifiedThe art-route layer of the Forge skills. route_media.py is the one command that generates an image or an image-to-video clip - through the paid API of any provider whose key is configured (OpenAI, Google Gemini, xAI, BytePlus ModelArk, fal.ai; the configured key is the owner's consent), else through the user's own signed-in Codex or Grok CLI - and prints one JSON line with the route, artifact, sha256 and cost estimate; with no route it answers no-route (codeart2d is then the last resort). Also the capability check (forge_doctor), API key configuration, the spend ledger, resume and Codex image adoption. Use when a Forge skill sends a generation here, for the capability check, or when the user asks about keys, routes or spend. Not for planning, processing or QA of sprites, video frames or maps (generate2dsprite, video2dsprite, generate2dmap) or for code-drawn art (codeart2d).
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 9.0 KB |
| agents/openai.yaml | config | 388 B |
| references/agent-profiles/README.md | markdown | 1.3 KB |
| references/agent-profiles/video-agent.md | markdown | 981 B |
| references/api-usage.md | markdown | 18.6 KB |
| references/capabilities.json | config | 32.9 KB |
| references/cli-routes.md | markdown | 15.3 KB |
| references/prices.json | config | 23.5 KB |
| references/provider-survey.md | markdown | 17.2 KB |
| references/route-media.md | markdown | 12.8 KB |
| references/schemas/common.schema.json | config | 12.9 KB |
| references/schemas/host.schema.json | config | 18.8 KB |
| references/schemas/media.schema.json | config | 22.5 KB |
| scripts/cli_media.py | script | 113.3 KB |
| scripts/forge_doctor.py | script | 74.5 KB |
| scripts/forge_index.py | script | 24.8 KB |
| scripts/generate_media.py | script | 72.1 KB |
| scripts/media_config.py | script | 9.1 KB |
| scripts/media_ledger.py | script | 29.8 KB |
| scripts/media_mp4.py | script | 9.4 KB |
| scripts/media_providers.py | script | 66.9 KB |
| scripts/route_media.py | script | 41.7 KB |
references/schemas/common.schema.json
View on GitHub{
"$schema": "https://json-schema.org/draft/2020-12/schema",
"$id": "https://raw.githubusercontent.com/0x0funky/agent-sprite-forge/main/shared/schemas/common.schema.json",
"title": "Agent Sprite Forge shared contract types",
"description": "Frozen contracts, version 1. Building blocks referenced by the sprite, video, map, codeart and media schemas. Canonical copy in shared/schemas; tools/vendor_sync.py copies it byte for byte into skills/*/references/schemas. Objects stay open (additionalProperties true): producers may add optional fields. Pixel boxes are half-open [x0, y0, x1, y1); paths are manifest-relative POSIX.",
"$defs": {
"sha256": {
"description": "Lowercase hexadecimal SHA-256 digest.",
"type": "string",
"pattern": "^[0-9a-f]{64}$"
},
"relPath": {
"description": "Manifest-relative POSIX path: forward slashes only, no leading slash, no drive letter and no URL scheme. '..' segments are allowed. Inputs on another drive record their file name plus sha256.",
"type": "string",
"minLength": 1,
"pattern": "^(?!/)(?![A-Za-z][A-Za-z0-9+.-]*:)[^\\\\]+$"
},
"fileRef": {
"description": "A file the document depends on, bound to its content by sha256.",
"type": "object",
"required": ["path", "sha256"],
"properties": {
"path": {"$ref": "#/$defs/relPath"},
"sha256": {"$ref": "#/$defs/sha256"},
"bytes": {"type": "integer", "minimum": 0}
}
},
"toolInfo": {
"description": "A tool or renderer and its exact version.",
"type": "object",
"required": ["name", "version"],
"properties": {
"name": {"type": "string", "minLength": 1},
"version": {"type": "string", "minLength": 1}
}
},
"timestamp": {
"description": "RFC 3339 date-time with an explicit offset, for example 2026-10-05T04:16:00Z.",
"type": "string",
"format": "date-time",
"pattern": "^[0-9]{4}-[0-9]{2}-[0-9]{2}T[0-9]{2}:[0-9]{2}:[0-9]{2}(\\.[0-9]+)?(Z|[+-][0-9]{2}:[0-9]{2})$"
},
"qaStatus": {
"description": "Overall verdict of a QA document. needs-visual-review means the numbers passed but a human or agent must still look.",
"enum": ["pass", "fail", "warn", "needs-visual-review"]
},
"qaCheck": {
"description": "One measured check. value and threshold are whatever the check measures (number, list, boolean or null). An ungated measurement is a skipped check with its value and a null threshold. Checks stay open: tools add keys such as subject (the file a per-file check is about; pixel_reduce), files (the files it names; palette_tool build), failing (the files or frames over the gate), profile (the lint profile), detail (extra figures such as loop_seam's seam and adjacent_median) and note (autotile_build).",
"type": "object",
"required": ["id", "status"],
"properties": {
"id": {"type": "string", "minLength": 1},
"status": {"enum": ["pass", "fail", "warn", "needs-visual-review", "skipped"]},
"value": true,
"threshold": true
}
},
"qaEnvelope": {
"description": "Every QA JSON states its method, what it does not prove, and the sha256 of every input and output it judged. A pass envelope cannot contain a failed check. createdAt is optional so QA files can stay byte-deterministic. Envelopes stay open: a tool adds its own top-level figures beside the envelope keys (for example palette_tool's palette, fit, totals, locked_inputs and luts, pixel_qa's frames and palette, autotile_build's metrics, parallax_build's seams, sweep and layers).",
"type": "object",
"required": ["status", "method", "notProven", "checks", "inputs", "outputs", "tool"],
"properties": {
"status": {"$ref": "#/$defs/qaStatus"},
"method": {"type": "string", "minLength": 1},
"notProven": {"type": "array", "items": {"type": "string", "minLength": 1}},
"checks": {"type": "array", "items": {"$ref": "#/$defs/qaCheck"}},
"inputs": {"type": "array", "items": {"$ref": "#/$defs/fileRef"}},
"outputs": {"type": "array", "items": {"$ref": "#/$defs/fileRef"}},
"tool": {"$ref": "#/$defs/toolInfo"},
"createdAt": {"$ref": "#/$defs/timestamp"}
},
"if": {"properties": {"status": {"const": "pass"}}, "required": ["status"]},
"then": {"properties": {"checks": {"items": {"properties": {"status": {"not": {"const": "fail"}}}}}}}
},
"artSource": {
"description": "Where the art came from. Record a separate boolean placeholder for stand-in art.",
"enum": ["host_image", "api", "code", "existing", "video", "mixed"]
},
"provenance": {
"description": "How an output was produced: the tool, its version, the parameters and the hashed inputs.",
"type": "object",
"required": ["tool", "version", "params", "inputs"],
"properties": {
"tool": {"type": "string", "minLength": 1},
"version": {"type": "string", "minLength": 1},
"params": {"type": "object"},
"inputs": {"type": "array", "items": {"$ref": "#/$defs/fileRef"}},
"renderer": {"$ref": "#/$defs/toolInfo"}
}
},
"point2": {
"description": "A finite [x, y] coordinate.",
"type": "array",
"items": {"type": "number"},
"minItems": 2,
"maxItems": 2
},
"size2": {
"description": "A [width, height] in whole pixels.",
"type": "array",
"items": {"type": "integer", "minimum": 1},
"minItems": 2,
"maxItems": 2
},
"box": {
"description": "Half-open pixel box [x0, y0, x1, y1) in the PIL convention: x1 and y1 are exclusive.",
"type": "array",
"items": {"type": "number"},
"minItems": 4,
"maxItems": 4
},
"rectXYWH": {
"description": "Rectangle [x, y, width, height] with its top-left corner at (x, y).",
"type": "array",
"prefixItems": [{"type": "number"}, {"type": "number"}, {"type": "number", "minimum": 0}, {"type": "number", "minimum": 0}],
"minItems": 4,
"maxItems": 4
},
"padding4": {
"description": "Padding [left, top, right, bottom] in whole pixels.",
"type": "array",
"items": {"type": "integer", "minimum": 0},
"minItems": 4,
"maxItems": 4
},
"polygon": {
"description": "A closed polygon as at least three [x, y] vertices; the last vertex joins the first.",
"type": "array",
"items": {"$ref": "#/$defs/point2"},
"minItems": 3
},
"hexColor": {
"description": "#RRGGBB, or #RRGGBBAA where AA 00 marks an erasing palette entry.",
"type": "string",
"pattern": "^#[0-9a-fA-F]{6}([0-9a-fA-F]{2})?$"
},
"rgb": {
"description": "An [r, g, b] colour with 8-bit channels.",
"type": "array",
"items": {"type": "integer", "minimum": 0, "maximum": 255},
"minItems": 3,
"maxItems": 3
},
"color": {
"description": "A colour as a hex string or an [r, g, b] triple.",
"anyOf": [{"$ref": "#/$defs/hexColor"}, {"$ref": "#/$defs/rgb"}]
},
"keyColor": {
"description": "A chroma-key colour: a declared key name or an explicit colour.",
"anyOf": [{"enum": ["magenta", "green", "blue"]}, {"$ref": "#/$defs/color"}]
},
"eventName": {
"description": "Gameplay event names shared by clips, selections and runtimes; anything else uses the custom: prefix.",
"anyOf": [
{"enum": ["in", "tell", "hit", "active_end", "cancel", "chain", "impact", "hold", "end", "sfx", "step_l", "step_r"]},
{"type": "string", "pattern": "^custom:\\S+$"}
]
},
"fpsValue": {
"description": "Frame rate as a positive number or an exact rational string such as 24000/1001.",
"anyOf": [
{"type": "number", "exclusiveMinimum": 0},
{"type": "string", "pattern": "^[1-9][0-9]*/[1-9][0-9]*$"}
]
},
"durationsMs": {
"description": "Per-frame durations in whole milliseconds (forge_core.frame_durations sums exactly to the clip length).",
"type": "array",
"items": {"type": "integer", "minimum": 1},
"minItems": 1
},
"loopPolicy": {
"description": "cycle repeats from the first frame, pingpong mirrors without duplicating end frames, oneshot plays once and holds.",
"enum": ["cycle", "pingpong", "oneshot"]
},
"sampling": {
"description": "Runtime texture sampling: nearest for pixel art, linear for painted art.",
"enum": ["nearest", "linear"]
},
"shadow": {
"description": "Ground shadow ellipse in source pixels, centred on the root.",
"type": "object",
"required": ["rx", "ry"],
"properties": {
"rx": {"type": "number", "minimum": 0},
"ry": {"type": "number", "minimum": 0},
"opacity": {"type": "number", "minimum": 0, "maximum": 1}
}
},
"seamReport": {
"description": "forge_core.seam_report of a loop, for QA envelopes of loop tools: seam is the premultiplied RGBA mean absolute difference (0-255) of the wrap (last -> first) and adjacent_* describe the steps between neighbours. seam_over_median and seam_over_p95 near 1 mean a seamless loop, well above 1 a pop at the wrap, near 0 a held duplicate frame; denominators are floored at 1e-6, so a static clip with a pop gives a very large ratio. Measure the decoded runtime file, inside the motion mask for scene loops; pass/fail thresholds belong to the adopting tool.",
"type": "object",
"required": ["seam", "adjacent_median", "adjacent_p95", "seam_over_median", "seam_over_p95"],
"properties": {
"frames": {"type": "integer", "minimum": 2},
"seam": {"type": "number", "minimum": 0},
"adjacent_median": {"type": "number", "minimum": 0},
"adjacent_p95": {"type": "number", "minimum": 0},
"adjacent_max": {"type": "number", "minimum": 0},
"seam_over_median": {"type": "number", "minimum": 0},
"seam_over_p95": {"type": "number", "minimum": 0},
"method": {"type": "string", "minLength": 1}
}
},
"edgeSeam": {
"description": "forge_core.edge_seam_report of one join or wrap: the one seam metric of every tool that judges edges (per D9: platform-strip joins, terrain wraps and Wang joins, parallax repeat seams). The common seamReport of the premultiplied RGBA steps (0-255) across the join (seam) and of the interior steps near it (adjacent_*), plus seam_ratio, near_median and a verdict. seam_ratio is the larger of seam / adjacent_max and the worst few-row window of the join over the worst window of the interior steps near it: about 1 or less looks like the art, well above 1 is a seam; method states the windows. Verdicts (snake_case): continuous (the join steps like the art around it), seam (a visible jump), duplicate_edge (a join much flatter than its neighbourhood: a repeated edge, a one-pixel stutter that edge equality would pass), flat (no measurable steps near the join, so there is nothing to compare) and too_small (fewer than two columns or rows to compare; the numbers may then be absent). A statistic, never a seamlessness proof; pass/fail gates belong to the adopting tool.",
"type": "object",
"required": ["verdict"],
"properties": {
"verdict": {"enum": ["continuous", "seam", "duplicate_edge", "flat", "too_small"]},
"seam_ratio": {"type": "number", "minimum": 0},
"near_median": {"type": "number", "minimum": 0}
},
"if": {"properties": {"verdict": {"const": "too_small"}}, "required": ["verdict"]},
"else": {"allOf": [{"$ref": "#/$defs/seamReport"}], "required": ["seam_ratio", "near_median"]}
},
"cropBoxes": {
"description": "forge-crop-boxes/v1 (DOC-18): the one form of measured crop boxes in sheet pixels for every tool. Each item is a half-open integer box [x0, y0, x1, y1) inside the sheet with an optional id. extract_prop_pack writes it as auto-boxes.json (with source, display_name and anchor_px in box pixels) and reads it with --boxes-file, where every item needs an id (the prop label) and may also carry footprint, solid, contact, occlusion_class and occupant_policy as in a prop pack item; assemble_frames --crop-boxes reads it with the id optional. Readers also accept the legacy forms: a bare list of boxes (assemble_frames) and {props: [{label, source_box}]} (extract_prop_pack).",
"type": "object",
"required": ["items"],
"properties": {
"schema": {"const": "forge-crop-boxes/v1"},
"source": {"$ref": "#/$defs/fileRef"},
"items": {
"type": "array",
"minItems": 1,
"items": {
"type": "object",
"required": ["box"],
"properties": {
"id": {"type": "string", "minLength": 1, "pattern": "\\S"},
"box": {"type": "array", "items": {"type": "integer", "minimum": 0}, "minItems": 4, "maxItems": 4},
"display_name": {"type": "string", "minLength": 1},
"anchor_px": {"description": "The prop's anchor in box pixels.", "$ref": "#/$defs/point2"}
}
}
}
}
}
}
}
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |