Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/serejaris/kimi-skillsnpx agentmods add skills/serejaris/kimi-skills/image_generationWrote 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/serejaris/kimi-skills/image_generation)<a href="https://agentmods.dev/skills/serejaris/kimi-skills/image_generation"><img src="https://agentmods.dev/badge/skills/serejaris/kimi-skills/image_generation/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/serejaris/kimi-skills/image_generation"><img src="https://agentmods.dev/badge/skills/serejaris/kimi-skills/image_generation.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00266 | $0.01810 |
| Opus 5 | $0.00133 | $0.00905 |
| Sonnet 5 | $0.00053 | $0.00362 |
| Haiku 4.5 | $0.00027 | $0.00181 |
Grade A, and why
image_generation scanned grade A with 1 finding 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 11d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
downloads the image to your `output` path with `curl` (the extension is How it starts
The opening of the file, as written. The whole thing — 159 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Image Generation
Use this skill to create an image from a text description with AI image generation, then save it locally and display it to the user.
Setup
Before the first use, ensure the agent-gw Python SDK (version 0.2.6 or newer) is installed. This checks the current environment and installs or upgrades it only when needed:
python3 scripts/image_generation_tool.py ensure-deps
The SDK needs an API key from api_key=..., KIMI_API_KEY, or
~/.kimi/agent-gw.json.
Parameters
description(required): detailed text description of the image to generate.ratio: one of1:1, 3:2, 2:3, 16:9, 9:16. Default1:1.resolution: one of1K, 2K, 4K. Default1K. Opaque background supports only these combinations:1K:1:1(1024x1024),3:2(1536x1024),2:3(1024x1536);2K:1:1(2048x2048),16:9(2048x1152);4K:16:9(3840x2160),9:16(2160x3840). Transparent background only supports1Kwith1:1, 3:2, 2:3.background:opaque(default) ortransparent.reference_image: public URL(s) that guide the generation. Repeat--reference-imagefor multiple. The gateway only accepts public URLs, so a local image must be converted withimage-to-urlfirst (see "Reference images" below); passing a local path togenerateis rejected.output(required): local output path ending in.jpg,.jpeg, or.png. Transparent background must use.png.
Workflow
- Build a detailed, descriptive
descriptionfrom the user's request: include subject, style, composition, lighting, colors, and mood. - Choose
ratio,resolution, andbackground. Enforce the supported combinations. For opaque images:1Kallows1:1 / 3:2 / 2:3;2Kallows1:1 / 16:9;4Kallows16:9 / 9:16. Transparent only allows1:1 / 3:2 / 2:3ratios,1Kresolution, and PNG output. - Pick an
outputfile path with a matching extension. - If the user supplies reference images, they must be public URLs. For any
local image (a file on the execution environment, whether that is a sandbox
or the client's local machine), first run the
image-to-urlcommand to upload it and get a public URL, then pass that URL with--reference-image. The gateway only accepts public reference URLs; passing a local path togenerateis rejected. - Run the
generatecommand (see "Script"). It callsgenerate_imageon the gateway, readsmedia.url/media.mime_typefrom the response, and downloads the image to youroutputpath withcurl(the extension is corrected to matchmime_type). - If the call fails, explain the failure reason from the printed error. Do not invent an image or a local path.
- On success, the script prints the saved file path. Then display the image
to the user by calling the
readFiletool on that path. Reading the image to show it is the model's job, not this plugin's work.
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.
- 11d ago First seen · 159 lines · 266 tokens per session scan A e01546db3a49
image_generation is a skill published in the GitHub repository serejaris/kimi-skills (6 stars, last pushed 1mo ago), licensed MIT. It adds 266 tokens to every session and 1,810 once invoked, about $0.0013 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
ape-present
Converts a blog post into a single self-contained HTML document -- a visual-heavy, readable walkthrough with custom animated diagrams, short and crisp bullet points, and a natural explanation flow from first principles. Trigger on "ape present", "make this presentable", or "turn this post into a doc I can walk people…
animation-reverse-engineering
Reverse-engineer any motion reference (a video from X/Twitter, Dribbble, a screen recording, a GIF) into production animation code through frame-level dissection. Use when the user shares a video/URL and says "implement this animation", "recreate this motion", "port this interaction", "how does this animate", "clone…
brand-guidelines
Use when applying or documenting a brand identity. Covers extracting a system from existing assets, documenting colour, type, and voice, and applying a brand consistently across new work.
generative-art
Use when creating algorithmic or generative visual art. Covers composition through code, controlled randomness, colour systems, and building work that is varied without being arbitrary.
visual-design
Use when creating posters, covers, social graphics, or other static visual designs. Covers composition, typographic hierarchy, colour, and producing original work rather than pastiche.
presentations
Use when creating or editing slide decks (.pptx). Covers slide structure, using layouts and templates correctly, charts and images, speaker notes, and building a deck that communicates rather than decorates.