Phase 2 — Stage → Launch

This skill turns a validated build sheet into exact API payloads and a resumable BYOK curl launch script, then guides the user through launching a Claude Managed Agent (environment → agent → session → kickoff) using their own Anthropic API key. It generates four ordered payloads, writes a launch script that reads the key from the environment, and validates everything before launch.

✨ What it does

  • Generates four ordered API payloads (environment, agent, session, kickoff) from a build sheet.
  • Writes a resumable launch.sh that reads $ANTHROPIC_API_KEY at runtime and never embeds it.
  • Runs a pre-launch validation including an API-key-leak scan that FAILs on any leak.
  • Enforces sequential launch order and resumability via .id files.
  • Provides a forcing-question library for key safety, environment choice, MCP permissions, and first poll monitoring.
  • No tool makes network calls; the user runs launch.sh themselves.

🎯 When to use it

  • When the user says 'launch it', 'deploy the agent', or 'create the agent now'.
  • When the orchestrator routes to phase=stage-launch.
  • After a build sheet has been validated and you need to produce runnable artifacts.
  • When you need to launch a Claude Managed Agent with the user's own API key without exposing it.

🚀 How to use

Trigger by saying 'launch it', 'deploy the agent', or 'create the agent now', or when the orchestrator routes to phase=stage-launch. Provide the path to a validated build sheet (e.g., ./my-agent/build-sheet.json). The skill runs three scripts: payload_generator.py, launch_script_writer.py, and payload_validator.py. Then the user must export their ANTHROPIC_API_KEY in their shell and run launch.sh. Example prompts:

Launch the agent from ./my-agent/build-sheet.json
Deploy the agent now using the build sheet in ./my-agent

📄 Output: Four JSON payload files, a resumable launch.sh script, and validation results.

📦 Add this skill to Claude Code

# 1. Get the skills repo
git clone --depth 1 https://github.com/alirezarezvani/claude-skills /tmp/claude-skills

# 2. Copy this skill into your project (or ~/.claude/skills for all projects)
mkdir -p .claude/skills
cp -r /tmp/claude-skills/agent-launcher/skills/stage-launch .claude/skills/stage-launch

Skill source: agent-launcher/skills/stage-launch/SKILL.md

⚠️ Good to know

Requires the user to have an Anthropic API key and run launch.sh themselves; the skill does not make network calls.

❓ FAQ

Does the skill ever handle the API key?

No. The key never enters chat, files, payloads, or logs. launch.sh reads it from the environment, and payload_validator.py scans for leaks.

What happens if I re-run launch.sh?

It resumes from the last created ID; each step skips if its .id file exists.

What order does launch.sh create resources?

It creates environment → agent → session → kickoff in that order, chaining IDs.

🤖 Overview, features, install steps and FAQ were generated from the project's SKILL.md on Sep 4, 2026. Always check the original source before running commands.