agentleFS
Sign inStart free
RoentekRoentek/Claude_Code_Boilerplate_Framework/.claude/skills/generate/SKILL.md

generate

Skill4 starsChanged 6 months agoMITView on GitHub

Generate images and videos via AI model APIs (Kie AI, with fal.ai and WaveSpeed AI as optional future providers). Routes each request to the cheapest capable model for the task, or an advanced model for high-fidelity work. Triggers on /generate, generate image, generate video, create image, create video, thumbnail, animate.

  • Reads credentials

19 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 4 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 17 edited versions we read keep these.

  • 4

    Distinctive

    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
  1. Distinctive:

    Read the recipe file before every generation — it has the exact model id, endpoint, request shape, and cost.

    line 20
  2. Distinctive:

    Never hide a provider swap.

    line 28
  3. Very specific:

    After every save, write the sidecar log: {basename}.json next to the media file (same rules as generations/README.md).

    line 35
  4. Very specific:

    Never describe a logo or face in text.

    line 41
  5. Distinctive:

    Run multiple generations one at a time to avoid rate limits.

    line 42
  6. Distinctive:

    If a request would generate more than a handful of items (e.g.

    line 43

“Common” is measured: kept by at least 6 in 10 of the 17 edited versions of this file we could read. “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

/generate

Playbook for AI media generation. Reads this file, picks a model, generates, saves flat, logs.

Models

TaskTierModelRecipe
Image, everyday/draftcheapNano Banana 2 Litemodels/nano-banana-2-lite.md
Image, text-in-image (signs, posters, UI)advancedGPT Image 2models/gpt-image-2.md
Video, general/defaultcheapKling 3.0models/kling-3.0.md
Video, higher quality from a start frameadvancedVeo 3.1models/veo-3.1.md
Video, animate reference imagesadvancedSeedance 2.0 Fastmodels/seedance-2.0-fast.md

Read the recipe file before every generation — it has the exact model id, endpoint, request shape, and cost.

Tier selection: default to the cheap tier (Nano Banana 2 Lite / Kling 3.0) for drafts, iteration, and simple requests. Move to the advanced tier only when the task needs it: readable text in the image (GPT Image 2), a hero-quality video shot (Veo 3.1), or animating 2-9 reference images instead of a text prompt (Seedance 2.0 Fast). If unsure which tier fits, say so and ask rather than guessing toward the expensive option.

Provider routing

  1. Kie AI is the only wired provider right now — use kie-cli (CLI-first, zero token overhead) or the kie-ai MCP tools as fallback. Key: KIE_AI_API_KEY (already in .claude/settings.local.json).
  2. fal.ai and WaveSpeed AI are documented as future fallbacks (see models/*.md provider notes) but are not yet wired — no FAL_KEY / WAVESPEED_API_KEY exist in this project. If a model isn't available on Kie AI, say so and ask before doing anything else; don't silently invent a fal.ai/WaveSpeed call.
  3. Never hide a provider swap. Say which route ran and why.

Output

  • Save every file FLAT into generations/ (project root). No subfolders.
  • Reference images live in generations/refs/.
  • Naming: {project}_{description}_{timestamp}.{ext}
  • After every save, write the sidecar log: {basename}.json next to the media file (same rules as generations/README.md).

Rules

  • Quote the cost and wait for explicit go-ahead before any paid video run. One approval = one run.
  • Draft on the cheap tier first. Only rerun on the advanced tier once a favorite is picked.
  • Never describe a logo or face in text. Pass the real image file from generations/refs/ as a reference. If it's missing, stop and ask for it.
  • Run multiple generations one at a time to avoid rate limits.
  • If a request would generate more than a handful of items (e.g. "generate 100 images"), stop and confirm the count and total cost before running any of it.

Styles

generations/styles.json holds reusable style presets (prompt fragment + reference file paths), editable from the gallery's Styles panel or by hand. When the user references a saved style by name, load its prompt fragment and refs from there instead of re-describing the look from scratch.

Gallery

generations/gallery-server.mjs serves a local gallery at the media library — see generations/README.md for how to launch it. The gallery reads generations/ directly; no manual import step, files just need to land in that folder per the Output rule above.

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.

Check your setup, free