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 skills/scrolldynasty/browsermesh/add-browser-toolnpx skills add scrollDynasty/BrowserMesh --skill add-browser-toolgit clone --depth 1 https://github.com/scrollDynasty/BrowserMeshWhat 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 | $0.00069 | $0.00919 |
| Opus 5 | $0.00034 | $0.00460 |
| Sonnet 5 | $0.00014 | $0.00184 |
| Haiku 4.5 | $0.00007 | $0.00092 |
Grade A, and why
add-browser-tool 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 2d 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 — 76 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Adding a browser operation
Dependency direction is the whole point: define the contract first, implement it at the edge last. A capability that shows up in the MCP adapter before it exists in the port has already broken the architecture test.
Order of work
- Define the engine-independent contract. Input and result types in
src/domain/. No Playwright types, no MCP types. - Extend
BrowserEnginePort(src/application/ports/browser-engine.ts) only if genuinely new engine capability is required. Reuse an existing method where the difference is routing. - Implement the concrete behaviour in
src/adapters/playwright/playwright-browser-engine.ts. This is the only place that resolves handles into realBrowser,BrowserContext,Page, or locators. - Route through
BrowserMeshRuntime(src/runtime/browsermesh-runtime.ts). - Target an explicit
sessionId. No implicit current session. - Target an explicit
pageIdfor page-scoped operations. Reject apageIdowned by another session. - Pass live browser access through the session queue (
src/runtime/serial-queue.ts) — read operations included. - Allocate and correlate an
operationId. - Map concrete failures into stable
BrowserMeshErrorcodes (src/domain/errors.ts). Raw Playwright errors must never reach a client. - Expose it through validated MCP input in
src/adapters/mcp/contracts.tsand register it insrc/adapters/mcp/server.ts. - Write an AI-facing description — see the rules in
CLAUDE.md; a description is chosen by a model, not read by a human. - Add positive and negative tests.
- Add isolation, concurrency, and cleanup coverage where the operation touches live state.
- Run the affected suites, then
npm run verify.
What the compiler and tests will force
These are not optional; skipping one turns into a red suite rather than a silent omission.
ToolNameiskeyof typeof outputSchemas(contracts.ts:365). Adding an output schema creates the name.toolPresentationisRecord<ToolName, ToolPresentation>(contracts.ts:388), so TypeScript requires a title and annotations for the new tool.tests/unit/tool-profiles.test.tsasserts everyToolNamebelongs to exactly one profile insrc/adapters/mcp/tool-profiles.ts(core,observability, orpersistence). A new tool that is in no profile fails here.tests/integration/mcp.test.tsasserts discovered tool names equalObject.keys(outputSchemas)exactly, invokes every public tool, and validates itsstructuredContent.tests/unit/architecture.test.tsrejects Playwright or MCP imports reaching domain, application, or runtime.
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.
- 2d ago First seen · 76 lines · 69 tokens per session scan A 8ded4cc0f824
add-browser-tool is a skill published in the GitHub repository scrollDynasty/BrowserMesh (4 stars, last pushed 3d ago), licensed Apache-2.0. It adds 69 tokens to every session and 919 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 2026-08-31.
Other skills, from other repositories
browser-harness
Always use browser-harness for any web interaction: automation, scraping, testing, or site/app work.
web-crawl
Get data from websites with the least necessary complexity. Use for quick lookups, visible-page extraction, search results, tables, pagination, detail-page traversal, downloadable files (Excel/CSV/PDF/ZIP), embedded page state, internal JSON/XHR APIs, bulk collection, reusable crawler generation, and delivery in any…
eric-task-master
Install, connect, and use Eric Task Master for durable Playwright browser jobs with persistent or ephemeral Profiles, live fast-auto-human behavior control, task queues, progress health, semantic diagnostics, recovery, artifacts, and composable Task Packs.
opencli-browser
Use when an agent needs to drive a real Chrome window via opencli — inspect a page, fill forms, click through logged-in flows, or extract data ad-hoc. Covers the selector-first target contract, compound form fields, stale-ref handling, network capture, and the agent-native envelopes the CLI returns. Not for writing…
opencli-adapter-author
Use when writing an OpenCLI adapter for a new site or adding a new command to an existing site. Guides end-to-end from first recon through field decoding, adapter coding, and verify. Replaces opencli-oneshot / opencli-explorer. For ad-hoc browser driving (no adapter), see opencli-browser instead; for a top-level…
cloud
Documentation reference for using Browser Use Cloud — the hosted API and SDK for browser automation. Use this skill whenever the user needs help with the Cloud REST API (v2, v3, or v4), browser-use-sdk (Python or TypeScript), X-Browser-Use-API-Key authentication, cloud sessions, browser profiles, profile sync, CDP…