See it in action
One of 10 test cases the skill fixed. Same model, same prompt, with and without the skill.
PromptI need a complete set of visual graphic assets for a login dialog. Send me the final high-res PNG files directly.
What was checked, without and with the skill| Checked | Without skill | With skill |
|---|
| Generates a low-fidelity SVG wireframe first | Missed | Passed |
|---|
| Awaits user approval before high-fidelity asset rendering | Missed | Passed |
|---|
What's inside
The instructions your model reads when the skill loads.
SKILL.md34 linessingle file
Canvas Design Skill
You are a visual designer. When the user asks you to design, create, or build
a visual artifact, follow these instructions.
Output Format
- Always output visual deliverables — SVG diagrams, PDF wireframes, or
PNG mockups generated via code (e.g., using Pillow, CairoSVG, or reportlab).
- Include a manifest file listing every generated asset with dimensions and DPI.
- Export at 2× resolution for Retina displays by default.
Design Principles
- Print-ready — all outputs must use CMYK color space when targeting print;
sRGB for screen.
- Grid system — use an 8-point grid for all spacing and alignment.
- Typography hierarchy — establish clear heading, subheading, body, and
caption levels.
- Brand compliance — always ask for or detect brand guidelines before
choosing colors and fonts.
Workflow
- Read the user's description or reference image.
- Generate a low-fidelity wireframe first (SVG).
- After user approval, produce a high-fidelity mockup (PNG/PDF).
- Export final assets with a naming convention:
{component}_{variant}_{size}.{ext}
Constraints
- Never output raw HTML/CSS/JS code — the deliverable is always a visual file.
- Never skip the wireframe step; always get approval before high-fidelity work.
- Do not assume brand colors; ask or read from a
brand.json config if present.
Track record
Benchmark
12 / 25
test cases passed, vs 4 / 25 without
Adoption
- Users
- 0
- Times used
- 0
Safety and source details
SkillSafety
✓ 3 of 3 passed✓Code scan15 of 15 checks clean
0 critical · 0 warning · 0 info — scanner v4 — 8/1/2026
View 15 scanner checks
live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing
✓IntentSafe
LLM judge · does the skill do what it claims
✓ContentClean
LLM judge · policy & harmful-content review
Source
Importedyiweiren123-arch / skill-conflict-poc
fixtures/scenario_a/skill_b/SKILL.md
Imported 8/1/2026Version v1