agentmods.dev

agentbrush: Instructions file for Claude Code

CLAUDE.md

agentbrush CLAUDE.md is an instructions file for Claude Code from ultrathink-art/agentbrush. It costs 523 tokens per session, scanned A, original, MIT.

Project instructions for AgentBrush, a tool that removes backgrounds from images, including rules for demonstration images and handling green areas trapped inside artwork.

In plain words
What is it for?
Use them when maintaining AgentBrush’s background-removal pipeline or updating its README examples.
Why use it?
They prevent common background-removal errors, such as damaging dark edges or leaving green patches behind.

Instructions file for Claude Code

Written for Claude Code: the file is CLAUDE.md.

This is ultrathink-art/agentbrush's own configuration. It tells Claude Code how to work on agentbrush itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything agentbrush configures →

not rated 9repo 3mo ago A scan 523 tokens original MIT
View source ↗ ultrathink-art/agentbrush
Reuse

Borrowing it

Nothing to install: this file belongs to ultrathink-art/agentbrush. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/ultrathink-art/agentbrush/main/CLAUDE.md
Clone the repo
git clone --depth 1 https://github.com/ultrathink-art/agentbrush

Made for: Claude Code.

Wrote 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.

agentmods badge for agentbrush CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/ultrathink-art/agentbrush/claude-md/github.svg)](https://agentmods.dev/instructions/ultrathink-art/agentbrush/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/ultrathink-art/agentbrush/claude-md"><img src="https://agentmods.dev/badge/instructions/ultrathink-art/agentbrush/claude-md/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.

agentmods 80×15 button for agentbrush CLAUDE.md

Your own site · 80×15
<a href="https://agentmods.dev/instructions/ultrathink-art/agentbrush/claude-md"><img src="https://agentmods.dev/badge/instructions/ultrathink-art/agentbrush/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 523 This file is loaded in full into every session.
When invoked 523 The same file — it is already loaded in full.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00523 $0.00523
Opus 5.5 $0.00209 $0.00209
Sonnet 5.5 $0.00105 $0.00105
Haiku 4.5 $0.00052 $0.00052

Measured 2mo ago against content hash 9aeeaa8f4316, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

agentbrush CLAUDE.md 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 2mo 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.

CLAUDE.md · 28 lines

How it starts

The opening of the file, as written. The whole thing — 28 lines — stays where its author put it; the contents beside it link to each section on GitHub.

README Visual Examples

  • BG removal demos: ALWAYS use WHITE background — white has zero color overlap with any colorful illustration. Black bg is risky: dark feather/fur/shadow details near edges are close to black threshold, flood fill nibbles them. Three iterations proved this. Current showcase: colorful owl on white bg.
  • Comparison images built with /tmp/rebuild_readme_examples.py pattern: Pillow checker-bg composites, side-by-side panels, labeled with find_font() Menlo fallback.
  • Composite example (#4) reuses the bg-removal cutout — if #1 looks bad, #4 cascades.

Trapped Green Screen (0.4.0)

  • Edge flood fill can never reach green enclosed by artwork — the sweep pass is the only thing that ever removed those pockets, and it only matches pixels still close to pure green. A pocket whose blend ring has drifted (or that a later hue/lift pass brightens) survives every pass and prints as a coloured bar. Detection is a REACHABILITY question, not a colour question.
  • Never assume the key colour. Real defect: art was keyed on (107,195,21) while the pipeline default was (24,242,41) — 83 apart on red, so key matching silently matched nothing. sample_key_color() reads it off the border.
  • Key mode gates, chroma mode only warns. Green-dominance + saturation is all that survives once the key has been blended away, and it also matches deliberate green artwork — measured 43% hit rate on a real 207-file catalogue. A gate that fires on 43% of shipped work gets ignored, so heuristic findings are warnings and the hard gate needs a key (--source / --key).
  • Labelling is run-based (horizontal runs + union-find), not per-pixel: ~0.5s on a 24MP file. numpy is required — per-pixel Python is not viable at that size.
  • Size thresholds on print files must be relative, not absolute. 40px is a real defect on a 1024px concept and invisible speckle on a 4500x5400 tee. Measured separation: real trapped field 91 ppm of canvas, sub-visible edge fringe 2.3 ppm — the default sits at 8 ppm with a 40px floor.

Read the full file on GitHub · 28 lines

Changes

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.

  1. 2mo ago First seen · 28 lines · 523 tokens per session scan A 9aeeaa8f4316

Subscribe to this mod's changes

agentbrush CLAUDE.md is an instructions file published in the GitHub repository ultrathink-art/agentbrush (9 stars, last pushed 3mo ago), licensed MIT. It adds 523 tokens to every session, about $0.0021 per session on Opus 5.5. 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.

Related

agentbrush CLAUDE.md compared with instructions from other repositories

agentbrush CLAUDE.md next to 6 instructions from other repositories
ModTypeStarsLast pushLicenceScanTokensWorks with
agentbrush CLAUDE.mdultrathink-art/agentbrushInstructions file93mo agoMITA523Claude Code
vscode buildNext.instructions.mdmicrosoft/vscodeInstructions file195ktodayMITA9,719GitHub Copilot
next.js AGENTS.mdvercel/next.jsInstructions file143ktodayMITA7,461Codex, OpenCode
langchain AGENTS.mdlangchain-ai/langchainInstructions file149ktodayMITA4,680Codex, OpenCode
vscode oss-third-party-notices.instructions.mdmicrosoft/vscodeInstructions file195ktodayMITA5,516GitHub Copilot
vscode copilot-instructions.mdmicrosoft/vscodeInstructions file195ktodayMITA3,093GitHub Copilot
deepseek-harness AGENTS.mddeepseek-ai/deepseek-harnessInstructions file275k2d agoMITA4,313Codex, OpenCode