safeclaw CLAUDE.md

safeclaw CLAUDE.md is an instructions file for coding agents from ykdojo/safeclaw. It costs 865 tokens per session, scanned A, original, MIT.

Development instructions for SafeClaw, a Claude Code environment running inside an isolated Docker container and accessed through a web terminal.

In plain words
What is it for?
Testing SafeClaw end to end, starting separate work or research sessions, and mounting local project folders for use inside the container.
Why use it?
They provide repeatable steps for rebuilding, testing, running multiple sessions, and connecting local projects to the container.

Instructions file

Install

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.

agentmods
npx agentmods add instructions/ykdojo/safeclaw/claude-md
Clone the repo
git clone --depth 1 https://github.com/ykdojo/safeclaw

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 safeclaw CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/ykdojo/safeclaw/claude-md.svg)](https://agentmods.dev/instructions/ykdojo/safeclaw/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/ykdojo/safeclaw/claude-md"><img src="https://agentmods.dev/badge/instructions/ykdojo/safeclaw/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 865 This file is loaded in full into every session.
When invoked 865 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
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 $0.00865 $0.00865
Opus 5 $0.00432 $0.00432
Sonnet 5 $0.00173 $0.00173
Haiku 4.5 $0.00086 $0.00086

Measured 4d ago against content hash d98b8b85c94b, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

safeclaw 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 4d 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 · 99 lines

How it starts

The opening of the file, as written. The whole thing — 99 lines — stays where its author put it; the contents beside it link to each section on GitHub.

SafeClaw

Sandboxed Docker container running Claude Code, accessible via a web terminal.

See architecture.md for full design details.

Testing end to end

After making changes, rebuild and test:

./scripts/build.sh
./scripts/run.sh

This opens http://localhost:7681 in the browser. Verify:

  1. Claude Code launches automatically with bypass permissions
  2. Confirm it shows the correct model (Opus 5) and doesn't ask for login
  3. Send a message and confirm it gets a response

Multiple sessions

Run multiple isolated sessions with -s:

./scripts/run.sh -n                    # default on port 7681
./scripts/run.sh -s work -n            # safeclaw-work on next available port
./scripts/run.sh -s research -n        # safeclaw-research on next available port

Mounting local projects

Use -v to mount a local directory into the container. Keep the same folder name inside the container for clarity:

./scripts/run.sh -s myproject -n -v /path/to/myproject:/home/sclaw/myproject

This mounts the project at /home/sclaw/myproject inside the container. Use the same folder name (not a generic "project") so it's clear which project you're working with. If the container already exists, it will be recreated with the new mount.

Research sessions

For web research or any task requiring URL fetching, use a SafeClaw container instead of doing it directly on the host. The -q option sends a query directly to Claude Code inside the container:

./scripts/run.sh -s research -n -q "Research Inngest and explain how durable execution works"

This starts the container (or reuses an existing one) and sends the query to Claude Code running inside it.

Dashboard

Start the dashboard to manage all sessions:

npx nodemon dashboard/server.js

Always use nodemon during development for auto-restart on changes.

Opens at http://localhost:7680. Shows all sessions with:

  • Start/stop/delete buttons
  • Live iframes of active sessions
  • Auto-refreshes via Docker events (SSE)

Read the full file on GitHub · 99 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. 4d ago First seen · 99 lines · 865 tokens per session scan A d98b8b85c94b

Subscribe to this mod's changes

safeclaw CLAUDE.md is an instructions file published in the GitHub repository ykdojo/safeclaw (181 stars, last pushed 6d ago), licensed MIT. It adds 865 tokens to every session, about $0.0043 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.