Claude Dev Setup
WarnAudited by ClawScan on May 10, 2026.
Overview
This is a coherent Claude Code setup helper, but it tells agents to read your Claude OAuth token and run Claude Code with permission checks disabled, including in background sessions.
Install only if you are comfortable letting this skill access your Claude Code OAuth token and run Claude Code tasks with permission checks disabled. Before use, remove or avoid `--dangerously-skip-permissions` unless you explicitly need it, do not paste or log raw tokens, review `setup.sh` before running it, and monitor or kill background sessions that are no longer needed.
Findings (5)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Claude Code may be able to modify files or take other coding actions without its normal permission prompts, so a mistaken or overly broad task could affect the project unexpectedly.
The documented default command runs Claude Code with `--dangerously-skip-permissions` for arbitrary task descriptions, including examples that run in the background.
CLAUDE_CODE_OAUTH_TOKEN=$(cat ~/.claude/.credentials.json | grep -o '"accessToken":"[^"]*"' | cut -d'"' -f4) claude --print --dangerously-skip-permissions 'Task description'
Do not make `--dangerously-skip-permissions` the default. Require explicit user approval for high-impact tasks, scope work to a specific project directory, and prefer normal Claude Code permission prompts.
Your Claude account token could be exposed in command logs, process listings, or agent context, and it grants delegated access to your Claude Code account.
The skill directs the agent to read a local OAuth credential file and extract the access token for every Claude Code call, while the registry metadata declares no credential requirement.
Claude Code stores its OAuth token in: ~/.claude/.credentials.json ... ALWAYS extract and use the token when calling Claude Code
Use the official Claude Code authentication flow where possible, avoid printing or inline-shelling raw tokens, declare the credential requirement clearly, and rotate the token if it is exposed.
Coding tasks may continue after the immediate response, consuming account resources or changing files until they are completed or killed.
The skill intentionally supports long-running background Claude Code sessions and provides a registry and kill/status workflow for them.
exec({
command: "claude --print --dangerously-skip-permissions 'Task description'",
background: true,
yieldMs: 10000
})Only start background sessions after explicit user intent, keep the session registry current, and regularly kill or clean up stale sessions.
Task labels, session IDs, and status notes persist across tasks; inaccurate or manipulated entries could affect future status checks or task decisions.
The skill creates a persistent markdown session registry that future agent actions read and update.
Before starting any Claude Code task, read this file to: 1. Check if a similar task is already running 2. Get the session ID for status queries 3. Clean up stale entries
Do not store secrets in the session file, keep entries minimal, and periodically review or clear old session records.
Running the setup script can change the local development environment and will trust the current npm package release.
The optional setup script installs a global npm package without a pinned version if the Claude CLI is missing.
npm install -g @anthropic-ai/claude-code
Review the script before running it, install from trusted package sources, and consider pinning or manually installing the Claude Code CLI version you intend to use.