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 agentmods add instructions/salmonbird/evoui-browser/agents-mdgit clone --depth 1 https://github.com/Salmonbird/evoui-browserWrote 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/instructions/salmonbird/evoui-browser/agents-md)<a href="https://agentmods.dev/instructions/salmonbird/evoui-browser/agents-md"><img src="https://agentmods.dev/badge/instructions/salmonbird/evoui-browser/agents-md.svg" alt="Measured on agentmods" 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 | $0.01099 | $0.01099 |
| Opus 5 | $0.00549 | $0.00549 |
| Sonnet 5 | $0.00220 | $0.00220 |
| Haiku 4.5 | $0.00110 | $0.00110 |
Grade A, and why
evoui-browser AGENTS.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 4d 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 — 114 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Project
Evoui Browser is an optional enhancement Skill for agent-browser. It records
browser operations, replays verified mechanical steps, and distills reusable
knowledge from completed tasks. The Agent always decides what to do next.
skills/evoui-browser/SKILL.md is the operating contract that an Agent follows
when using the Skill. The runtime implementation lives in
skills/evoui-browser/scripts/evoui/. This file defines the rules for changing
this repository; see README.md for installation and usage.
Public repository layout
evoui-browser/
├── .github/workflows/ci.yml
├── skills/evoui-browser/
│ ├── SKILL.md
│ ├── agents/openai.yaml
│ ├── references/{agent-browser,runtime,knowledge,evolution}/
│ └── scripts/
│ ├── evoui.py
│ └── evoui/
├── tests/
├── .gitignore
├── AGENTS.md
├── README.md
├── SECURITY.md
├── LICENSE
└── NOTICE
This tree lists only stable entries in the public repository. Git-ignored local data and development material are intentionally omitted.
Product principles
- Stay decoupled from
agent-browser; pass through by default. Without--evoui, Evoui must preserve the original invocation exactly: argv, standard streams, TTY behavior, signals, and exit code must match a directagent-browserinvocation. - Accelerate repeated operations; do not make business decisions. Evoui executes browser operations whose intent is already known. The Agent chooses the target, interprets the page, and decides what to do next.
- Keep Replay scoped. Evoui does not try to replay every action in a task. Partial Replay is useful when it removes mechanical work and cleanly returns uncertain work to the Agent. Replay must stop when the page does not match or the result cannot be verified. A click or submission that may already have taken effect must not be retried mechanically. Making unstable or judgment-heavy actions replayable is outside the product scope.
- Success does not imply that a task is worth learning. Knowledge should be saved only when the scope is clear, the steps are likely to recur, and the result can be verified.
- Do not hard-code product exceptions for individual websites. Site, selector, and product-specific differences belong in scoped knowledge, not in product code.
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.
- 4d ago First seen · 114 lines · 1,099 tokens per session scan A c55549aac199
evoui-browser AGENTS.md is an instructions file published in the GitHub repository Salmonbird/evoui-browser (102 stars, last pushed 20d ago), licensed Apache-2.0. It adds 1,099 tokens to every session, about $0.0055 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 instructions, from other repositories
pi-agent-browser-native AGENTS.md
Instructions for fitchmultz/pi-agent-browser-native, covering pi-agent-browser-native, product-specific guidance, documentation placement, upstream capability baseline and command reference and tool result categories.
obscura AGENTS.md
AGENTS.md instructions for h4ckf0r0day/obscura, covering agents.md, build, rendering and stealth, no rendering, with rustls or stealth and test.
browser-harness CLAUDE.md
Claude Code instructions for browser-use/browser-harness, a project described as: Browser Harness | Self-healing harness that enables LLMs to complete any task.
browser-control AGENTS.md
Instructions for keon/browser-control: Use this project as a tiny browser pipe, not an agent framework.
stealth-browser-mcp AGENTS.md
Instructions for vibheksoni/stealth-browser-mcp, covering agent instructions, start here, recommended proxy services, environment and validation.
actionbook CLAUDE.md
Instructions for actionbook/actionbook, covering claude.md, project overview, architecture, domain model hierarchy and module structure.