Parselbox: Instructions file for Claude Code

CLAUDE.md

Parselbox CLAUDE.md is an instructions file for Claude Code from thesanjeetc/Parselbox. It costs 987 tokens per session, scanned A, original, MIT.

Project instructions for Parselbox, a secure Python sandbox that lets AI agents run code and access connected services through controlled interfaces.

In plain words
What is it for?
Use them when modifying the sandbox, its host-to-sandbox communication, service bridges, execution tool, or documentation.
Why use it?
They explain the project’s architecture, conventions, commands, and known pitfalls so changes fit the existing system.

Instructions file for Claude Code

Written for Claude Code: the file is CLAUDE.md. Also seen: reads .claude/ paths; mentions CLAUDE.md.

This is thesanjeetc/Parselbox's own configuration. It tells Claude Code how to work on Parselbox itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything Parselbox configures →

Reuse

Borrowing it

Nothing to install: this file belongs to thesanjeetc/Parselbox. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/thesanjeetc/Parselbox/main/CLAUDE.md
Clone the repo
git clone --depth 1 https://github.com/thesanjeetc/Parselbox

Made for: Claude Code.

Wrote 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.

agentmods badge for Parselbox CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/thesanjeetc/parselbox/claude-md/github.svg)](https://agentmods.dev/instructions/thesanjeetc/parselbox/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/thesanjeetc/parselbox/claude-md"><img src="https://agentmods.dev/badge/instructions/thesanjeetc/parselbox/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.

agentmods 80×15 button for Parselbox CLAUDE.md

Your own site · 80×15
<a href="https://agentmods.dev/instructions/thesanjeetc/parselbox/claude-md"><img src="https://agentmods.dev/badge/instructions/thesanjeetc/parselbox/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 987 This file is loaded in full into every session.
When invoked 987 The same file — it is already loaded in full.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00987 $0.00987
Opus 5 $0.00494 $0.00494
Sonnet 5 $0.00197 $0.00197
Haiku 4.5 $0.00099 $0.00099

Measured 9d ago against content hash 54b5d0c25557, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

Parselbox 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.

CLAUDE.md · 45 lines

How it starts

The opening of the file, as written. The whole thing — 45 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

Parselbox: a secure Python sandbox for AI agents — one execute_code MCP tool; external systems (MCP servers, REST/GraphQL, shells, host functions) appear as native Python objects inside it. Single process: Deno + Pyodide, no containers. It is an execution layer that plugs into any harness — harness-scope features (planning, memory, orchestration) are out of scope.

Architecture (host ↔ sandbox)

Host (Python, holds credentials)                Deno subprocess (permission-jailed)
  main.py      Parselbox orchestrator             sandbox/main.ts     entry, registers "exec"
  rpc.py       JSON-lines RPC over stdio    <-->  sandbox/rpc.ts      counterpart
  bridge/      tools: http, graphql, shell,       sandbox/sandbox.ts  PyodideManager (mounts, timeout,
               mcp (OAuth), toolkit (sbx.*)                           packages, watcher, restart)
  mcp.py       exposes execute_code (FastMCP)     sandbox/serve.ts    @api routes, uploads, live reload
  view.py      display() → MCP Apps renderer      sandbox/wasi.ts     WASI preview1 shim — .wasm commands
  context.py   dispatches callbacks to bridges                        via require() and as bash() commands
  models.py    Callback / ExecutionResult / Mount sandbox/filesystem.ts  just-bash (pure-JS bash) + watcher
  prompt.py    agent guide (served by sbx.help()) sandbox/setup.ts    Pyodide boot + snapshot cache
  hooks.py     lifecycle + MCP elicitation        sandbox/bootstrap.py  runs INSIDE Pyodide: sbx toolkit,
                                                                      namespace proxies, tasks, js/require/bash

Two RPC flows: host→sandbox exec (run code), sandbox→host callback (a namespace proxy call → bridge executes with real creds → result returns). Tool calls made in sandbox code therefore run on the host; secrets never enter the sandbox.

Commands

  • uv sync — setup (Deno ≥2.x must be installed separately)
  • make test — coverage + pytest over tests/ (excludes bench); boots real sandboxes, ~7 min
  • uv run pytest tests/test_foo.py -q — run one file; make bench — benchmarks
  • make check — the CI gate: deno lint + ruff + ruff format --check, then make test. CI runs the same lint/format hooks via pre-commit.

Read the full file on GitHub · 45 lines

Changes

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.

  1. 9d ago First seen · 45 lines · 987 tokens per session scan A 54b5d0c25557

Subscribe to this mod's changes

Parselbox CLAUDE.md is an instructions file published in the GitHub repository thesanjeetc/Parselbox (4 stars, last pushed 12d ago), licensed MIT. It adds 987 tokens to every session, about $0.0049 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.

Related

Other instructions, from other repositories