What is ClawCrate CLAUDE.md?

Repository instructions for ClawCrate, a Rust runtime that runs AI-agent shell commands inside operating-system sandboxes without Docker, virtual machines, or root access.

Use them when working on command execution, sandbox profiles, automatic language detection, workspace copies, secret filtering, or platform support.

They describe how profiles, temporary workspaces, environment scrubbing, and platform-specific restrictions fit together, helping changes preserve command isolation.

How manuelpenazuniga/ClawCrate configures Claude Code

Reuse

Borrowing it

Nothing to install: this file belongs to manuelpenazuniga/ClawCrate. 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/manuelpenazuniga/ClawCrate/main/CLAUDE.md

Clone the repo

git clone --depth 1 https://github.com/manuelpenazuniga/ClawCrate

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.

README.md

[![agentmods](https://agentmods.dev/badge/instructions/manuelpenazuniga/clawcrate/claude-md/github.svg)](https://agentmods.dev/instructions/manuelpenazuniga/clawcrate/claude-md)

Your own site

<a href="https://agentmods.dev/instructions/manuelpenazuniga/clawcrate/claude-md"><img src="https://agentmods.dev/badge/instructions/manuelpenazuniga/clawcrate/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.

Your own site · 80×15

<a href="https://agentmods.dev/instructions/manuelpenazuniga/clawcrate/claude-md"><img src="https://agentmods.dev/badge/instructions/manuelpenazuniga/clawcrate/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 8,733 This file is loaded in full into every session.

When invoked 8,733 The same file — it is already loaded in full.

Security scan C 2 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.

Model	Per session	Once invoked
Fable 5.1	$0.08733	$0.08733
Opus 5.5	$0.03493	$0.03493
Sonnet 5.5	$0.01747	$0.01747
Haiku 4.5	$0.00873	$0.00873

Measured 2mo ago against content hash df6c3a8b3839, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade C, and why

ClawCrate CLAUDE.md scanned grade C with 2 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 2mo 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.

Reaches for credential fileshighPrivilege escalation

SSH keys, cloud credentials, git-credentials, .npmrc, /etc/shadow: reading these is how a config file becomes a credential leak.

- Tests (Linux only): fixture that tries `cat ~/.ssh/id_rsa` → EACCES. Fixture that tries `strace` → EPERM.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

│ ├── exfiltration_attempt/ # Python script trying urllib to evil.com

CLAUDE.md · 744 lines

How it starts

The opening of the file, as written. The whole thing — 744 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md — ClawCrate Development Guide

Project Overview

ClawCrate is a secure execution runtime for AI agents. A single Rust binary per platform (~15-20MB) that sandboxes shell commands using native OS primitives: Landlock + seccomp on Linux, Seatbelt on macOS. No Docker, no VMs, no root required.

This file is the source of truth for all development work. Read it fully before writing any code.

Specification document: docs/clawcrate-v3.1.1.md contains the full project specification including architecture decisions, competitive landscape, and rationale for every design choice. This CLAUDE.md extracts the actionable implementation details.

Core Architecture

clawcrate run --profile build -- cargo test
    │
    ├── Parse CLI args (clap)
    ├── Resolve profile (safe|build|install|open or custom YAML)
    ├── Auto-detect stack if no profile specified (Cargo.toml → rust, package.json → node)
    ├── Determine workspace mode: DefaultMode from profile + CLI flags (--replica/--direct)
    ├── Materialize WorkspaceMode: if Replica, create temp dir + filtered copy
    ├── Scrub environment variables (remove *_SECRET*, *_TOKEN, *_KEY, etc.)
    ├── Generate execution plan (plan.json)
    │
    ├── LAUNCH (platform-specific):
    │     Linux:  fork → set rlimits → apply Landlock → apply seccomp → exec
    │     macOS:  generate SBPL → exec via /usr/bin/sandbox-exec
    │
    ├── CAPTURE: pipe stdout/stderr, monitor child process
    ├── COLLECT: wait for exit, compute fs-diff (snapshot pre vs post)
    ├── WRITE ARTIFACTS: plan.json, result.json, stdout.log, stderr.log, audit.ndjson, fs-diff.json
    │
    └── Report result to user (human-readable or --json)

Key Design Decisions — DO NOT DEVIATE

Deny by default. The sandboxed process starts with zero permissions and receives only what the profile grants. Never start permissive and restrict.
Platform-native sandboxing only. Linux uses Landlock + seccomp. macOS uses Seatbelt. No Docker, no VMs, no containers. The sandbox is kernel-level, irremovible from inside, and inherited by all child processes.

Read the full file on GitHub · 744 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.

2mo ago First seen · 744 lines · 8,733 tokens per session scan C df6c3a8b3839

Subscribe to this mod's changes

ClawCrate CLAUDE.md is an instructions file published in the GitHub repository manuelpenazuniga/ClawCrate (2 stars, last pushed 3mo ago), licensed MIT. It adds 8,733 tokens to every session, about $0.0349 per session on Opus 5.5. A static security scan graded it C with 2 findings (reaches for credential files, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

AI Safety Building agents ai-agents ai-security audit-trail compliance developer-tools

ClawCrate CLAUDE.md next to 6 instructions from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
ClawCrate CLAUDE.mdmanuelpenazuniga/ClawCrate	Instructions file	2	3mo ago	MIT	C	8,733	Claude Code
vscode buildNext.instructions.mdmicrosoft/vscode	Instructions file	195k	today	MIT	A	9,719	GitHub Copilot
vscode oss-third-party-notices.instructions.mdmicrosoft/vscode	Instructions file	195k	today	MIT	A	5,516	GitHub Copilot
next.js AGENTS.mdvercel/next.js	Instructions file	143k	today	MIT	A	7,461	Codex, OpenCode
langchain AGENTS.mdlangchain-ai/langchain	Instructions file	149k	today	MIT	A	4,680	Codex, OpenCode
vscode copilot-instructions.mdmicrosoft/vscode	Instructions file	195k	today	MIT	A	3,093	GitHub Copilot
vscode model-prompts.instructions.mdmicrosoft/vscode	Instructions file	195k	today	MIT	A	2,022	GitHub Copilot

ClawCrate: Instructions file for Claude Code