start

A start command for launching ClaudeClaw in either background daemon mode or one-shot mode. It first checks that the project is not being initialized from the user's home folder and verifies required runtimes.

In plain words
What is it for?
Starting the heartbeat daemon, running one prompt or trigger, checking runtime availability, and previewing the required setup conditions.
Why use it?
It prevents an unsafe project location and detects missing Bun or Node.js before startup, reducing failed or misplaced installations.

Command

Part of the claudeclaw plugin — 3 skills, 10 commands shipped together

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 commands/moazbuilds/claudeclaw/start
Clone the repo
git clone --depth 1 https://github.com/moazbuilds/claudeclaw

Or install claudeclaw, the plugin that ships this one along with the rest of its 3 skills, 10 commands.

Per session 10 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,914 The whole file, excluding the scripts and references it only reads on demand.
Security scan D 3 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 $0.00010 $0.03914
Opus 5 $0.00005 $0.01957
Sonnet 5 $0.00002 $0.00783
Haiku 4.5 $0.00001 $0.00391

Measured 3d ago against content hash b9a48b630e43, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade D, and why

start scanned grade D with 3 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 3d 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.

Downloads and executes remote codehighSupply chain

curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.

curl -fsSL https://bun.sh/install | bash

Reads agent configuration directoriesmediumAgent snooping

.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.

Wait 1 second, then check `cat .claude/claudeclaw/logs/daemon.log`. If it contains "Aborted: daemon already running", tell the user and exit.

Makes network callslowCapability

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

curl -fsSL https://bun.sh/install | bash
commands/start.md · 263 lines

How it starts

The opening of the file, as written. The whole thing — 263 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Start the heartbeat daemon for this project. Follow these steps exactly:

  1. Block home-directory starts (CRITICAL, BLOCKER):

    • Run pwd and echo "$HOME".
    • If pwd equals $HOME, STOP immediately.
    • Tell the user exactly:
      • "CRITICAL BLOCKER: For security reasons, close this session and start a new one from the folder you want to initialize ClaudeClaw in."
    • Do not continue with any other step until they restart from a non-home project directory.
  2. Runtime checker (Bun + Node):

    • Run:
      which bun
      which node
      
    • If bun is missing:
      • Tell the user Bun is required and will be auto-installed.
      • Run:
        curl -fsSL https://bun.sh/install | bash
        
      • Then source the shell profile to make bun available in the current session:
        source ~/.bashrc 2>/dev/null || source ~/.zshrc 2>/dev/null || true
        
      • Verify again with which bun. If still not found, tell the user installation failed and to install manually from https://bun.sh, then exit.
      • Tell the user Bun was auto-installed successfully.
    • If node is missing:
      • Tell the user Node.js is required for the OGG converter helper.
      • Ask them to install Node.js LTS and rerun start, then exit.
  3. Check existing config: Read .claude/claudeclaw/settings.json (if it exists). Determine which sections are already configured:

    • Heartbeat configured = heartbeat.enabled is true AND heartbeat.prompt is non-empty
    • Telegram configured = telegram.token is non-empty
    • Discord configured = discord.token is non-empty
    • Security configured = security.level exists and is not "moderate" (the default), OR security.allowedTools/security.disallowedTools are non-empty
  4. Interactive setup — smart mode (BEFORE launching the daemon):

    If ALL three sections are already configured, show a summary of the current config and ask ONE question:

    Use AskUserQuestion:

    • "Your settings are already configured. Want to change anything?" (header: "Settings", options: "Keep current settings", "Reconfigure")

    If they choose "Keep current settings", skip to step 6 (first contact question). If they choose "Reconfigure", proceed to step 5 below as if nothing was configured.

    If SOME sections are configured and others are not, show the already-configured sections as a summary, then only ask about the unconfigured sections in step 5.

    If NOTHING is configured (fresh install), ask about all three sections in step 5.

  5. Ask setup questions:

    Use AskUserQuestion to ask all unconfigured sections at once (up to 3 questions in one call):

    • Model (always ask if model is empty/unset): "Which Claude model should ClaudeClaw use?" (header: "Model", options: "opus (default)", "sonnet", "haiku", "glm")
    • If heartbeat is NOT configured: "Enable heartbeat? Example: I can remind you to drink water every 30 minutes, or you can fully customize what runs." (header: "Heartbeat", options: "Yes" / "No")
    • If Telegram is NOT configured: "Configure Telegram? Recommended if you want it 24/7 live." (header: "Telegram", options: "Yes" / "No")
    • If Discord is NOT configured: "Configure Discord? Connect your bot to Discord servers." (header: "Discord", options: "Yes" / "No")
    • If security is NOT configured: "What security level for Claude?" (header: "Security", options:
      • "Moderate (Recommended)" (description: "Full access scoped to project directory")
      • "Locked" (description: "Read-only — can only search and read files, no edits, bash, or web")
      • "Strict" (description: "Can edit files but no bash or web access")
      • "Unrestricted" (description: "Full access with no directory restriction — dangerous"))

    Then, based on their answers:

    • Model: Set model in settings to their choice (e.g. "opus", "sonnet", "haiku", "glm"). Default is "opus" if they don't pick.
    • If model is glm: Ask in normal free-form text for API token and set top-level api to that value (optional; user can skip). Only ask this token question when the selected model is glm.

Read the full file on GitHub · 263 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. 3d ago First seen · 263 lines · 10 tokens per session scan D b9a48b630e43

Subscribe to this mod's changes

start is a command published in the GitHub repository moazbuilds/claudeclaw (1,313 stars, last pushed 1mo ago), licensed MIT. It adds 10 tokens to every session and 3,914 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it D with 3 findings (downloads and executes remote code, reads agent configuration directories, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.