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/starling-browser/starling/agents-mdgit clone --depth 1 https://github.com/starling-browser/starlingWrote 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/starling-browser/starling/agents-md)<a href="https://agentmods.dev/instructions/starling-browser/starling/agents-md"><img src="https://agentmods.dev/badge/instructions/starling-browser/starling/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 | $0.05796 | $0.05796 |
| Opus 5 | $0.02898 | $0.02898 |
| Sonnet 5 | $0.01159 | $0.01159 |
| Haiku 4.5 | $0.00580 | $0.00580 |
Grade A, and why
starling AGENTS.md scanned grade A with 1 finding 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 3d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
`curl http://localhost:8088/<name>/`. A site that renders from disk but not How it starts
The opening of the file, as written. The whole thing — 448 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agents — rules of engagement
This file is the first thing an implementation agent should read when opening this repo. It explains how multiple agents share work without stepping on each other, and how a single agent can stop mid-task and have a later session resume cleanly.
TL;DR
# 1. Orient
cat AGENTS.md # this file
less browser-plan/13_MILESTONES.md # where we are
less tasks/INDEX.md # what's available
# 2. Claim something unblocked
./tasks/lib/claim.sh claim wp:M1-03-dom-core "agent-claude-<your-handle>"
# 3. Read the package file and start working on main
less tasks/M1/wp-M1-03-dom-core.md
# 4. Commit often, with the wp id in the subject:
# "wp:M1-03 — Node hierarchy + tests"
# 5. Build + test before completing
dotnet build && dotnet test
# 6. Mark complete:
./tasks/lib/claim.sh complete wp:M1-03-dom-core
All work happens on main. This repo doesn't use a per-package branch
workflow — agents commit directly to main. (If a remote with PR review is
ever wired up, the optional in_review state in tasks/SCHEMA.md is
available for it; today it's unused.)
If you have to stop early, leave a handoff log entry in the task file
and either keep the claim (you'll resume) or release it
(./tasks/lib/claim.sh release wp:…). Either way: commit so the next
agent sees the state.
The contract
| You can rely on | You must do |
|---|---|
One file per work package under tasks/M*/wp-*.md |
Touch only your claimed package's file (plus your code changes) |
tasks/INDEX.md reflects the current status |
Update INDEX.md when you change a status |
Work happens on main; commits carry the wp id in the subject |
Prefix commit messages with wp:<id> — so history is greppable |
| Inbound contributions require a CLA | Make sure the contributor has a CLA on file before the work lands |
Dependencies are explicit in depends_on |
Don't start a package whose deps aren't complete |
| Stale claims age out at 72 h | Add a handoff log entry every session, even if you're "still going" |
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.
- 3d ago First seen · 448 lines · 5,796 tokens per session scan A 47538d393d1f
starling AGENTS.md is an instructions file published in the GitHub repository starling-browser/starling (57 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 5,796 tokens to every session, about $0.0290 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
electron CLAUDE.md
Instructions for electron/electron, covering electron development guide, running nodemodules binaries, project overview, directory structure and build tools setup.
electron copilot-instructions.md
Instructions for electron/electron, covering copilot instructions for electron, build system, linting, running a single test and example: npm run test -- -g "ipc".
html-video CLAUDE.md
Instructions for nexu-io/html-video, covering html-video 项目工作区, 角色与边界, 产品定位（决策时间线）, 关键差异化 vs hyperframes and 目录结构.
skelet AGENTS.md
Instructions for Selekkt/skelet, covering agents.md — developing skelet.css, what skelet is, source of truth, repo map and rules of engagement (framework dev).
snag AGENTS.md
Instructions for p3bot/snag, covering agents.md, project overview, setup commands, clone repository and install dependencies.
plasmate CLAUDE.md
Instructions for plasmate-labs/plasmate, covering plasmate, quick start, build, run and mcp mode (for claude code, cursor, etc.).