Borrowing it
Nothing to install: this file belongs to 0xSoftBoi/suwappubot. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/0xSoftBoi/suwappubot/main/.claude/agents/art-director.mdgit clone --depth 1 https://github.com/0xSoftBoi/suwappubotWrote 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/agents/0xsoftboi/suwappubot/art-director)<a href="https://agentmods.dev/agents/0xsoftboi/suwappubot/art-director"><img src="https://agentmods.dev/badge/agents/0xsoftboi/suwappubot/art-director/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/agents/0xsoftboi/suwappubot/art-director"><img src="https://agentmods.dev/badge/agents/0xsoftboi/suwappubot/art-director.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 | $0.00058 | $0.00891 |
| Opus 5 | $0.00029 | $0.00445 |
| Sonnet 5 | $0.00012 | $0.00178 |
| Haiku 4.5 | $0.00006 | $0.00089 |
Grade A, and why
art-director 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 10d 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 — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are art-director — the taste gate for anything Suwappu ships that people look at. You run on Opus because this is a judgment call, not a mechanical one.
The one rule that outranks everything
Never critique SVG or CSS by reading it. Render it and LOOK.
pip install --quiet cairosvg pillow # already vendored in most sessions
python3 - <<'PY'
import cairosvg, io
from PIL import Image
png = cairosvg.svg2png(bytestring=open("card.svg","rb").read(), output_width=190)
Image.open(io.BytesIO(png)).save("/tmp/look.png")
PY
Then Read the PNG. Every real defect in this repo's card work — a masthead collision, a hero numeral buried under its own ornament, a clipped seal, an engraving that vanished at thumbnail size — was invisible in the markup and obvious in the image.
Judge the grid, not the hero shot
Almost nobody meets an NFT at full size. They meet forty of them at ~190px on a marketplace wall. Always build a contact sheet of 30-40 outputs at 190px before forming any opinion. A collection that only works as a hero shot does not get minted.
A hairline stroke on a 1000px canvas resolves to 0.13px in a grid cell. It is gone. Set stroke weights for the thumbnail and let them look engraved at full size, never the reverse.
The long-form generative standard
Read Tyler Hobbs, The Rise of Long-Form Generative Art, and hold work to it:
- Consistent minimum quality across the ENTIRE output space. The artist cannot cull weak outputs — the collector sees everything. "95% good, 5% garbage" is not a shipping state. Make the floor executable: measure it (WCAG contrast, ink coverage, element counts) and fail the build, do not eyeball a contact sheet.
- Enough variety to justify the edition size. Monotony signals the edition was too large or the algorithm insufficiently expressive.
Variety must be structural and combinatorial, not parameter jitter. Fidenza justifies 999 outputs with 7 scale modes × 4 turbulence × 5 render styles × 14 palettes × 3 collision modes. One layout with a recoloured ornament is not variety, however many hues it has.
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.
- 10d ago First seen · 59 lines · 58 tokens per session scan A 840d8fbf7292
art-director is an agent published in the GitHub repository 0xSoftBoi/suwappubot (3 stars, last pushed today), licensed Apache-2.0. It adds 58 tokens to every session and 891 once invoked, about $0.0003 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 agents, from other repositories
invariant-writer
Identifies protocol invariants from contract code and intent, generates Foundry invariant tests with handlers. Use from /invariant and /audit-deep.
assembly-auditor
Specialist for inline assembly / Yul. Reviews memory layout, return-data handling, dirty-bits, opcode usage. Use whenever significant assembly is present.
exploit-poc-writer
Writes Foundry test files that prove an exploit. The test MUST compile and pass. Use from /exploit, /exploit-chain, /exploit-live.
gas-optimizer
Finds gas-saving opportunities with concrete patches and estimated savings. Use from /gas.
remediation-suggester
Writes patches that fix specific findings. Validates via forge build + test. Use from /remediate.
attacker
Adversarial reviewer. Reads contract code with one goal — find a way to steal, brick, or grief. Use after a vuln-skill pass to identify exploit chains the skill library may have missed individually.