fal-ai-generation
Skill800 starsChanged 23 days agoMITView on GitHub
Generate and edit images, create video, audio, 3D assets or material maps through fal.ai MCP or the Python queue client. Use for fal model discovery, reference uploads, generation, recovery and local delivery.
- Reads credentials
In this repository43 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .agents/skills/img2threejs/CLAUDE.mdCLAUDE.md
- CLAUDE.mdCLAUDE.md
- .claude/skills/img2threejs/CLAUDE.mdCLAUDE.md
- 3d-production-routingSkill
- blender-game-animationSkill
- character-sheet-pipelineSkill
- feature-handoffSkill
- game-animation-integrationSkill
- game-audio-workflowSkill
- game-level-designSkill
- gameplay-visual-reviewSkill
- game-ui-workflowSkill
- game-vfx-workflowSkill
- higgsfield-workflowSkill
- image-reference-workflowSkill
- img2threejsSkill
- materials-to-gameSkill
- motion-reference-workflowSkill
- planSkill
- video-watchSkill
- voice-dictation-setupSkill
- workspace-authoringSkill
- 3d-production-routingSkill
- blender-game-animationSkill
- character-sheet-pipelineSkill
- fal-ai-generationSkill
- feature-handoffSkill
- game-animation-integrationSkill
- game-audio-workflowSkill
- game-level-designSkill
- gameplay-visual-reviewSkill
- game-ui-workflowSkill
- game-vfx-workflowSkill
- higgsfield-workflowSkill
- image-reference-workflowSkill
- img2threejsSkill
- materials-to-gameSkill
- motion-reference-workflowSkill
- planSkill
- video-watchSkill
- voice-dictation-setupSkill
- workspace-authoringSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 6 are distinctive, and 2 are 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.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 9
Use the connected fal.ai MCP when available.
- Very specific: line 20
When no model is selected, call recommend_model with the actual task.
- Distinctive: line 26
Read MCP calls for parameter examples and recovery.
- Distinctive: line 27
Read prompting and review for concept, reference-edit and material decisions.
- Distinctive: line 38
Do not pass every model the same argument object.
- Very specific: line 40
Prefer submit_job for long operations.
- Distinctive: line 46
If submission returned no ID, inspect request history before trying again.
- Distinctive: line 52
Keep the prompt, source filenames or hashes, model ID, request ID, local outputs and review verdict together.
≈ “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.
Files
fal.ai generation
Turn the user's brief and supplied references into downloaded, reviewed assets. Use the connected fal.ai MCP when available. The included Python helper handles local uploads and queue recovery when MCP cannot access the local filesystem. This skill and its relative resources can be copied into another project.
Choose the operation
Distinguish concept exploration, faithful editing, reference-sheet preparation, motion generation, mesh generation and material creation. Preserve a selected provider, model, accepted image, pose, camera and requested output count. An edit changes the requested feature; it does not redesign the whole asset.
When no model is selected, call recommend_model with the actual task. Then read
get_model_schema for the exact endpoint. Similar display names on different
providers are not interchangeable endpoint IDs. Inspect pricing for the planned
batch, especially video or 3D; use existing task authorization without asking
again for already-approved work. Defaults in an old script are not a model choice.
Read MCP calls for parameter examples and recovery. Read prompting and review for concept, reference-edit and material decisions. Read setup when connecting another machine. Each recipient uses their own account and credentials.
Run and retain the job
- Inspect the selected input files. Upload only the references needed for the requested task. A hosted MCP cannot read a local file path: use the upload helper or a client-side upload tool, then pass the resulting URL.
- Prepare inputs from the live schema. Fields such as
image_urls,image_url,aspect_ratio,image_size,resolutionand duration differ by endpoint. Do not pass every model the same argument object. Do not add web search, background removal, extra views or video unless they serve this task. - Prefer
submit_jobfor long operations;run_modelis convenient for images. Immediately save endpoint, request ID, returned status/result URLs and the input parameters in the task directory.processingis normal, not failure. - Continue with
check_job, thenget_job_resultwhen complete. Prefer the canonical URLs returned by the service. A wait timeout does not authorize a duplicate submission. A terminal result can still contain a provider error. If submission returned no ID, inspect request history before trying again. - Download the outputs before relying on a CDN link. Inspect each requested asset, preserve rejected takes with their reasons, and make the chosen result easy to open and download. Record technical completion separately from visual acceptance. A generated mesh still needs its own rig/topology/export checks.
Keep the prompt, source filenames or hashes, model ID, request ID, local outputs and review verdict together. Credentials and raw signed links do not belong in a shareable report. In Mr. Mak, use the current Workspace card and its shared image viewer; elsewhere use the recipient's requested output directory.
Included tools and examples
- Queue helper:
--help, upload, submit, status and result download. One receipt per job; an existing receipt blocks another submission. - Python dependencies.
- Concept input and edit input: neutral requests, with no private artwork or working account URLs.
- Example task: a non-spending rehearsal followed by the commands for an explicitly requested generation.
Resolve these paths against this skill directory. Run --help or submit
--dry-run to check installation without a generation. Runtime keys come from
the environment or an explicitly selected .env; never print their values.
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
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.

