Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/devkindhq/ideogram-ai-toolkit/custom-model-trainingnpx skills add devkindhq/ideogram-ai-toolkit --skill custom-model-traininggit clone --depth 1 https://github.com/devkindhq/ideogram-ai-toolkitWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/devkindhq/ideogram-ai-toolkit/custom-model-training)<a href="https://agentmods.dev/skills/devkindhq/ideogram-ai-toolkit/custom-model-training"><img src="https://agentmods.dev/badge/skills/devkindhq/ideogram-ai-toolkit/custom-model-training.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00230 | $0.01472 |
| Opus 5 | $0.00115 | $0.00736 |
| Sonnet 5 | $0.00046 | $0.00294 |
| Haiku 4.5 | $0.00023 | $0.00147 |
Grade A, and why
custom-model-training scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 108 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Custom Model Training
Ideogram lets you train a custom model on your own reference images, then generate new
images that stay consistent with those references via a custom_model_uri passed to
generate_image. This is the skill that closes the loop other skills in this toolkit
start: brand-identity-sheet locks a brand system into one image; character-model-sheet
locks a character into one multi-panel turnaround. Once that reference exists, this
skill turns it (plus any other reference images) into a model that generates on-brand
or on-character assets indefinitely, instead of re-describing the same look in every
future prompt and hoping it stays consistent.
Always run the pipeline — create the dataset, upload the images, kick off training,
poll until it's ready, and generate a proof image with the trained model — rather than
stopping after train_model and telling the user to check back later. The prompt-only
version of this skill would just be a description of the Ideogram API; the value is in
actually running it, watching training through to completion, and coming back with a
generated image that demonstrates the model works.
Before you start: read the honest facts
Read references/dataset-requirements.md before running the pipeline. It splits what's
actually confirmed about these tools (from direct inspection of their schemas) from
what's genuinely unknown (minimum image count, training duration, the exact "ready"
status value). Don't invent numbers for the unknowns — tell the user what's confirmed
and what you're finding out by trying it, per the standing rule against stating
third-party API behavior as fact without a verified source.
Workflow
1. Resolve the input
Ask for (or confirm) a local folder of reference images if the user hasn't pointed to one already. This version of the skill only supports a folder of existing images — if the user wants to train on images generated earlier in this session, save those to a folder first, then proceed the same way.
What ships with it
27 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- evals/evals.json 6.0 KB
- examples/anchorpoint-logos/anchorpoint-logo-1.png 352 KB
- examples/anchorpoint-logos/anchorpoint-logo-2.png 352 KB
- examples/anchorpoint-logos/anchorpoint-logo-3.png 352 KB
- examples/anchorpoint-logos/anchorpoint-logo-4.png 352 KB
- examples/anchorpoint-logos/anchorpoint-logo-5.png 352 KB
- examples/fizzwright-hopcarry-training-set/01-fizzwright-identity-sheet.webp 48 KB
- examples/fizzwright-hopcarry-training-set/02-kip-hopcarry-character-sheet.webp 43 KB
- examples/fizzwright-hopcarry-training-set/03-kip-holding-fizzwright-bottle.webp 49 KB
- examples/fizzwright-hopcarry-training-set/04-fizzwright-bottle-product-shot.webp 33 KB
- examples/fizzwright-hopcarry-training-set/05-fizzwright-kip-delivery-box.webp 49 KB
- examples/fizzwright-hopcarry-training-set/06-kip-icon-fizzwright-style.webp 163 KB
- examples/fizzwright-hopcarry-training-set/07-kip-sprint-pose.webp 28 KB
- examples/fizzwright-hopcarry-training-set/08-fizzwright-wordmark-hero.webp 11 KB
- examples/fizzwright-hopcarry-training-set/09-fizzwright-icon-citrus-peel.webp 44 KB
- examples/fizzwright-hopcarry-training-set/10-fizzwright-icon-bottlecap.webp 26 KB
- examples/fizzwright-hopcarry-training-set/11-fizzwright-icon-bubbles.webp 19 KB
- examples/fizzwright-hopcarry-training-set/12-fizzwright-icon-cherry-glyph.webp 3.0 KB
- examples/fizzwright-hopcarry-training-set/13-kip-face-detail.webp 5.8 KB
- examples/fizzwright-hopcarry-training-set/14-kip-bicycle-delivery-poster.webp 8.6 KB
- examples/fizzwright-hopcarry-training-set/15-kip-satchel-bottles-detail.webp 7.7 KB
- examples/fizzwright-hopcarry-training-set/16-fizzwright-buttons-ui.webp 2.6 KB
- examples/fizzwright-hopcarry-training-set/kip-delivering-fizzwright-order.webp 34 KB
- examples/fizzwright-hopcarry-training-set/kip-surfing-bottlecap-wave.webp 40 KB
- examples/fizzwright-hopcarry-training-set/off-brand-steampunk-astronaut-test.webp 39 KB
- examples/fizzwright-hopcarry-training-set/RESULT.md 3.4 KB
- references/dataset-requirements.md 4.9 KB
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 108 lines · 230 tokens per session scan A 720e5266ad52
custom-model-training is a skill published in the GitHub repository devkindhq/ideogram-ai-toolkit (2 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 230 tokens to every session and 1,472 once invoked, about $0.0011 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
model-integration
Use when adding a new model or pipeline to diffusers, setting up file structure for a new model, converting a pipeline to modular format, or converting weights for a new version of an already-supported model.
custom-blocks
Use when the user has written (or wants to write) a ModularPipelineBlocks subclass in a local Python file and needs to package it into a Hub-uploadable directory. Covers the workflow from a single block.py file to a published custom-block repo that consumers can load via ModularPipeline.frompretrained( …
ima2
Use the ima2-gen CLI/server to generate, edit, inspect, and manage local AI image generation jobs.
videoagent-image-studio
Tired of juggling 8 API keys? This skill gives you one-command access to Midjourney, Flux, Ideogram, and more, with zero setup. Use when you want to generate any image without worrying about API keys.
flux-3-product-ads
Use when building a finished product ad from FLUX 3 - shot design, voiceover, action-to-word sync, evidence-gated copy, deterministic assembly, and QC gates that catch clipped audio, floating products, off-model plates, and reports that claim a pass the build did not give.
flux-3-prompt-doctor
Use when diagnosing a FLUX 3 brief before generation. Resolve missing, conflicting, or schema-changing requirements.