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/ilharp/webview-devtools-mcp/agents-mdgit clone --depth 1 https://github.com/ilharp/webview-devtools-mcpWhat 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 | $0.00761 | $0.00761 |
| Opus 5 | $0.00380 | $0.00380 |
| Sonnet 5 | $0.00152 | $0.00152 |
| Haiku 4.5 | $0.00076 | $0.00076 |
Grade A, and why
webview-devtools-mcp 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 yesterday.
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 — 38 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Changes to upstream forks live as stgit patches — keep diffs minimal
packages/chrome-devtools-mcp, packages/chii, and packages/chobitsu are forks of upstream projects (each is its own nested git repo). Our modifications to them are not ordinary commits — they are a stacked-git (stg) patch stack, and the exported patch files live under patches/<package>/ in the root repo (the series file records patch order).
Implication: every change you make to an upstream-originated file in these forks ends up as a hunk in some patch. Therefore:
- Make the smallest possible, logically-correct change. First inspect the baseline and your diff with
git -C packages/<pkg> show HEAD:<file>/git -C packages/<pkg> diff, and leave untouched any line you don't have to change. - Do not "clean up" pre-existing upstream dead code, duplication, or style. That bloats the patch, makes it dirty, and makes rebasing onto a newer upstream harder. Touch only what your feature actually needs.
- When you need a value for later use, prefer inlining or a minimal local extraction over adding large blocks near the top of the file (unless truly necessary).
- For the
packages/chobitsupackage, put newly-written utility functions underpackages/chobitsu/src/wdmcp/whenever possible, rather than modifying the original files. This way you can freely edit the code in the wdmcp folder without being bound by the minimal-change constraint.
Which files you can freely rewrite
Files added by our patches (not upstream files) are owned by this project — normal editing or full rewrites are fine. How to tell: git -C packages/chrome-devtools-mcp log --oneline -- <file>, or check which patch in patches/chrome-devtools-mcp/ introduces it. Already in upstream = edit minimally; added by a patch = edit freely.
The patch stack is managed by hand
- The stack is managed manually with the
stgcommand line. There are no scripts / automation / build hooks that apply or refresh patches. - Do not run
stg(push/pop/refresh/...),git commitinside the nested forks, or modify files underpatches/, unless the user explicitly asks. The maintainer folds working-tree changes into the patch stack themselves withstg. - Your job is usually: get the source into a correct, minimal shape and leave it in the working tree. Turning it into a patch is the maintainer's step.
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.
- yesterday First seen · 38 lines · 761 tokens per session scan A f2c8bc5e106e
webview-devtools-mcp AGENTS.md is an instructions file published in the GitHub repository ilharp/webview-devtools-mcp (2 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 761 tokens to every session, about $0.0038 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 2026-08-31.
Other instructions, from other repositories
surf-cli AGENTS.md
Instructions for nicobailon/surf-cli, a project described as: The CLI for AI agents to control Chrome. Zero config, agent-agnostic, battle-tested.
chrome-mcp-secure CLAUDE.md
Instructions for Pantheon-Security/chrome-mcp-secure, covering chrome mcp server - claude code integration, quick start, one-command setup (installs, builds, starts chrome), verify connection and available tools.
chrome-console-log-mcp-server AGENTS.md
AGENTS.md instructions for ONLYHUMN/chrome-console-log-mcp-server, a project described as: Capture browser console logs via the Chrome DevTools Protocol and expose them to an MCP client over SSE.
chrome-devtools-mcp AGENTS.md
AGENTS.md instructions for ChromeDevTools/chrome-devtools-mcp, covering instructions and rules for typescript.
dev3000 AGENTS.md
Instructions for vercel-labs/dev3000, covering agents.md, runtime, local ui, production workflows, browser tools and development rules.
neobrowser AGENTS.md
Instructions for pitiflautico/neobrowser, covering agents.md — orientation for ai agents & contributors, build, run, test (rust — primary), how it works (rust module map), conventions and ci and release gates.