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 skills add guinacio/claude-image-gen --skill character-reference-sheetsgit clone --depth 1 https://github.com/guinacio/claude-image-genWrote 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/guinacio/claude-image-gen/character-reference-sheets)<a href="https://agentmods.dev/skills/guinacio/claude-image-gen/character-reference-sheets"><img src="https://agentmods.dev/badge/skills/guinacio/claude-image-gen/character-reference-sheets.svg" alt="Measured on agentmods" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 18 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
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 | $0.00117 | $0.01325 |
| Opus 5 | $0.00059 | $0.00662 |
| Sonnet 5 | $0.00023 | $0.00265 |
| Haiku 4.5 | $0.00012 | $0.00133 |
Grade A, and why
character-reference-sheets 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 7d 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 — 118 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Character reference sheets
Pipeline: a nude body render comes out of Blender, the model dresses that body, each garment is cut out, it goes to an image-to-3D generator, and the mesh comes back into Blender.
The goal of the generation is not a beautiful image. It is the same body as the render, now clothed, with every piece legible and cuttable. Everything below exists to protect that.
Cost rule, before anything else
Never call create_asset without asking first. Every generation spends the
user's API credit. Write the full prompt, surface the decisions that change the
image, and wait for an explicit go-ahead. This applies to each retry after a bad
result too — do not silently fix and re-run.
Workflow
- Measure the input renders with
scripts/measure.py. If any criterion in section 8 ofreferences/resolutions-and-proportions.mdfails, say so and ask for a re-export from Blender. Cheaper than discovering the problem after the image is paid for. - Flatten alpha onto white if the PNG is RGBA. Alpha often becomes black on upload and contaminates the result.
- Build the prompt using the architecture in
references/prompt-template.md. - Ask for permission to generate.
- Generate, one view per call.
- Measure the output and compare against the input (thresholds below).
- Copy from
IMAGE_OUTPUT_DIRinto the character project's own output folder, with a descriptive versioned name:character_outfit_front_v2.png.
Measurement dependency
scripts/measure.py requires Python and Pillow. Before first use, ask before
installing the dependency, then run:
python -m pip install -r /path/to/character-reference-sheets/requirements.txt
If Pillow is unavailable and cannot be installed, do not generate: report that the mandatory pre-flight measurement could not be completed.
The rule that matters most
Never describe the body in the prompt. Not the build, not shoulder width, not head size, not height. Describing anatomy makes the model draw a new body from the text instead of preserving the one in the image.
What ships with it
5 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.
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.
- 7d ago First seen · 118 lines · 117 tokens per session scan A 42f72ac16b98
character-reference-sheets is a skill published in the GitHub repository guinacio/claude-image-gen (62 stars, last pushed 3d ago), licensed MIT. It adds 117 tokens to every session and 1,325 once invoked, about $0.0006 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 2026-08-30.
Other skills, from other repositories
image-to-3d-pipeline
Transformez une image 2D en modèle 3D animé prêt pour le web ou le jeu en moins de 30 minutes, en utilisant le workflow Dilum Sanjaya (Hunyuan3D + Mixamo). Use when: Créer un personnage 3D pour un site web - Mascotte, avatar, illustration interactive; Prototyper un asset de jeu - Character design, props…
screenshot-isolated
Render a target GameObject from a chosen camera angle with optional layer-based isolation, configurable background (solid/skybox/transparent), multi-light setup via JSON, and Composite (2x2 Front/Right/Back/Top) mode. Returns a PNG image. When isolated=true, inactive children may briefly fire OnEnable — see the body…
screenshot-camera
Capture a screenshot from a Unity Camera and return it as a PNG image for direct LLM inspection. Falls back to Camera.main (then any active camera) when cameraRef is null. Width and height are capped to keep response size manageable.
screenshot-game-view
Capture a screenshot of the Unity Editor's Game View by reading its internal render texture directly. Image size matches the current Game View resolution; the tool corrects Y-flip on DirectX / Metal so the output is always upright. Requires an open Game View window.
screenshot-scene-view
Capture a screenshot from the Unity Editor Scene View at the requested size. Renders via the Scene View's active camera onto a temporary RenderTexture. Requires an open Scene View.
sound-cues
Create and modify SoundCue assets — add/connect nodes (mixer, random, delay, attenuation, modulator, etc.) and set audio properties (SoundCueService). Use when the user asks to create a Sound Cue, build a SoundCue node graph, or wire audio playback logic.