agentsclimarketplace
Sign in

Image generation

Skill event4u-app/agent-config/src/skills/image-generation

Generate an image from a brief — provider-agnostic blueprint then provider-specific translation, with ref-image/seed reuse for consistency. Use when generating/creating an image.From its SKILL.md

Install
npx -y skills add event4u-app/agent-config --skill image-generation

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 7 stars7 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

5.0 KB, ~1.0k tokens by cl100k_base, as published. Nobody here has run it

image-generation

Generate an image end-to-end: capture the brief, route to the right provider, author a provider-specific prompt, reuse seeds for consistency, invoke the adapter, and apply governance. All adapters are scaffold-tier (dry-run) until promoted to stable.

When to use

  • User asks to generate, create, or produce an image from a brief.
  • End-to-end image production (routing + prompting + adapter invocation + governance).
  • When a consistent series of images needs seed/ref-image reuse across renders.

Procedure

  1. Capture the brief — extract: subject, output format (raster/vector/banner/icon), style intent, any existing ref images or seed values for consistency.
  2. Route the provider via image-provider-routing — match job shape (text-in-image → Ideogram, photoreal → Flux, vector → Recraft, general → Gemini/GPT Image 2).
  3. Author the prompt via prompt-engineering-image — apply provider-specific grammar (text-literal first for Ideogram, noun-phrase for Flux, style: param for Recraft, natural language for Gemini/GPT).
  4. Reuse ref-image / seed for consistency — if the brief is part of a series, carry the seed value or ref-image path forward. Seed reuse is the primary consistency lever; re-describing the subject each time is not.
  5. Invoke the adapter (dry-run today) — run src/scripts/ai-image/adapters/<provider>.sh with the assembled params. Validate the returned artifact path or dry-run confirmation. All adapters are experimental (scaffold-tier); no live generation occurs until a maintainer promotes the adapter via provider-lifecycle-discipline.
  6. Apply governance — run the rights check (image-likeness-and-rights) when the brief names a real person, brand mark, or living artist's style. Attach the AI-disclosure footer per media-governance-routing before delivering the output.

Output format

  1. Blueprint — provider choice + routing rationale + prompt string ready to copy.
  2. Provider + prompt — adapter file reference, key params (aspect ratio, style, negative prompts), and any seed/ref-image value carried forward.
  3. Artifact path / dry-run note — the path returned by the adapter, or an explicit note: "adapter is experimental (scaffold-tier) — dry-run plan only; no rendered asset until promotion per provider-lifecycle-discipline."

Gotcha

  • Scaffold-tier adapters produce plans, not pixels — all four adapters (ideogram.sh, flux.sh, recraft.sh, gemini-image.sh) are scaffold-tier (dry-run only). This skill produces the blueprint + dry-run confirmation; actual renders require a maintainer to capture a smoke trace and promote the adapter to stable. Claiming a rendered asset exists when no adapter is stable misleads the caller.
  • Seed reuse is the consistency lever — re-describing the subject more precisely in each prompt does not lock character or style; carrying the seed value or ref-image path forward does. Drop the seed and character drift is inevitable across a series.

Good example: Brief for a product-shot series → route to Flux (photoreal), author noun-phrase prompt, carry seed: 42 across all five renders, invoke adapter, note "experimental — dry-run plan only."

Bad example: Invoking the adapter and telling the user "here is your rendered image" while all adapters are still scaffold-tier.

Do NOT

  • Do NOT claim a rendered asset is produced while adapters are scaffold-tier — surface the dry-run caveat explicitly every time.
  • Do NOT skip image-likeness-and-rights when the brief names a real person, a brand mark, or a named living artist's style.
  • Do NOT bypass image-provider-routing — selecting the wrong provider for the job shape (e.g. Flux for a text-in-image logo) produces garbled output.
  • Do NOT ignore seed/ref-image values when consistency across a series is required — seed reuse is the only reliable consistency mechanism.

See also

What ships with it: 1 file

1.6 KB alongside SKILL.md

evals/

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.