Borrowing it
Nothing to install: this file belongs to zinin/sketchup-mcp2. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/zinin/sketchup-mcp2/master/CLAUDE.mdgit clone --depth 1 https://github.com/zinin/sketchup-mcp2Wrote 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/zinin/sketchup-mcp2/claude-md)<a href="https://agentmods.dev/instructions/zinin/sketchup-mcp2/claude-md"><img src="https://agentmods.dev/badge/instructions/zinin/sketchup-mcp2/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.
<a href="https://agentmods.dev/instructions/zinin/sketchup-mcp2/claude-md"><img src="https://agentmods.dev/badge/instructions/zinin/sketchup-mcp2/claude-md.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.03235 | $0.03235 |
| Opus 5 | $0.01618 | $0.01618 |
| Sonnet 5 | $0.00647 | $0.00647 |
| Haiku 4.5 | $0.00324 | $0.00324 |
Grade A, and why
sketchup-mcp2 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 9d 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 — 199 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
Guidance for Claude Code (claude.ai/code) working in this repository.
What This Project Does
MCP Server for SketchUp bridges Claude AI and SketchUp via the Model Context Protocol (MCP). Two components:
- Python MCP server (
src/sketchup_mcp/) — receives Claude's tool calls, forwards them over TCP - Ruby SketchUp extension (
mcp_for_sketchup/mcp_for_sketchup/) — TCP server inside SketchUp, executes commands against the live model
Non-Obvious Constraints
- Units: SketchUp's internal Ruby API uses inches; all MCP tools accept and return mm. Convert at the boundary via
MM = 25.4. Group#subtractis reversed:A.subtract(B)returnsB - A. To get «target minus tool», calltool.subtract(target). Verified empirically against SketchUp 2026.- SketchUp is single-threaded: the Ruby extension cannot use threads; all I/O runs in
UI.start_timercallbacks. - Wire protocol (v0.0.1+): 4-byte big-endian length-prefix framing, 64 MiB cap.
- Persistent socket: Python server holds one TCP connection;
asyncio.Lockserializes tool-calls. Ruby reads non-blocking insideUI.start_timer, capped at 50 reads per tick (~3.2 MB) to keep SketchUp's UI responsive. - Ruby supports N concurrent TCP clients:
core/server.rbkeeps@clients(sock → ClientState) + a global FIFO@frame_queue. Each timer tick: accept connections, drain reads into each client'sFrameReader, dispatch frames in FIFO decode-arrival order (accept-order across clients, decode-order within a client — single shared queue, not round-robin). Operations still serialise on the single-threaded SketchUp UI thread. Per-client errors close only that client. Logical races between clients on the model are the user's responsibility (no server-side locking). Half-open sockets detected viaSO_KEEPALIVE(OS default ~2h). - Entity IDs: SketchUp's
find_entity_by_idrequires Integer; cast incoming string IDs with.to_i. - Solid tools are unreliable on non-manifold geometry:
boolean_operationand edge ops use copy-based + sequential-per-edge workarounds. Sketchup::Model#undodoes not exist: programmatic undo dispatchesSketchup.send_action("editUndo:").- Request IDs round-trip: both sides preserve the JSON-RPC
idso async responses can be matched. - Mutating handlers wrap edits in
model.start_operation/commit_operationsoundorolls back atomically. - eval gate:
eval_rubyis gated by theeval_enabledpref, which ships on (Config::DEFAULTS[:eval_enabled] = true) and is closed fromPlugins → MCP Server → Settings.... An absent pref resolves to that default; a present-but-non-boolean pref fails closed — a corrupt value is no basis for enabling arbitrary code execution. Gate closed ⇒handlers/eval.rb::eval_rubyraises JSON-RPC-32010, whichtools.py::eval_rubyturns into a user-facing message (no[code]prefix) so the LLM repeats it verbatim. Arbitrary-code risk is guarded in two layers: (1) a blocking enable-time security confirm (ui/settings_dialog.rb::confirm_eval_enable— warns it grants full filesystem/network/shell access), shown on an off→on transition inside the Settings dialog — and only there. An upgrade does not pass through the dialog, so an install with no stored pref (never opened Settings) gets the new open default with no confirm, including one upgrading from a-warehousebuild where the gate was closed; (2) per-call review at the MCP client — Claude Desktop / Claude Code show eacheval_rubycall's code to approve or deny. Note this second layer is a client convention, not a protocol guarantee: MCP does not mandate an interaction model, and a client running pre-authorized (--dangerously-skip-permissions) shows nothing. The extension deliberately does NOT log or re-prompt the code — that would duplicate the client's permission UI and break autonomous (--dangerously-skip-permissions) operation. - Version handshake (one-time on connect): every TCP connection MUST
begin with a JSON-RPC
hellorequest carryingparams.client_version. The server validates againstcore/compat.rb'sMIN_PYTHON..MAX_PYTHONrange and replies with{server_version, client_id}inresult. Mismatches return JSON-RPC error-32001(IncompatibleVersionErroron the Python side) and the server closes the socket. After a successful handshake, regulartools/callrequests carry noclient_versionfield and responses carry noserver_versionfield. Compatibility ranges live insrc/sketchup_mcp/compat.pyandmcp_for_sketchup/mcp_for_sketchup/core/compat.rb.get_versionremains a regular tool returning the verdict payload. - Literal source-guard tests:
test/test_operation_names.rb,test/test_transform_absolute.rb,test/test_joints_frame_compensation.rbpin exact handler source text (down to indentation) to protect invariants like the reversedGroup#subtract. Never run an auto-formatter over handlers or tests; when refactoring pinned code, update the pins deliberately.
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.
- 9d ago First seen · 199 lines · 3,235 tokens per session scan A 51a6b799ac60
sketchup-mcp2 CLAUDE.md is an instructions file published in the GitHub repository zinin/sketchup-mcp2 (20 stars, last pushed 10d ago), licensed MIT. It adds 3,235 tokens to every session, about $0.0162 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
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
deepseek-harness AGENTS.md
AGENTS.md instructions for deepseek-ai/deepseek-harness, covering agents.md, pre-stable apis and released session data, repository layout, commands and host sandbox failures.