skills.rest
alirezarezvani
alirezarezvaniCOMMUNITY
Explore all skills by this author →
Website𝕏@RezaRezvaniBlnGitHub

stage-launch

Generate CMA API payloads and a resumable BYOK curl launch script from a build sheet.

25.3k|3.6k|Updated Oct 19, 2025
Search Tags:#agent deployment#byok#Claude Managed Agents#api key safety#api payloads#cURL#launch script
One-click install
npx skills add https://github.com/alirezarezvani/claude-skills --skill stage-launch
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: stage-launch
Source: https://github.com/alirezarezvani/claude-skills/tree/main/agent-launcher/skills/stage-launch
Command: npx skills add https://github.com/alirezarezvani/claude-skills --skill stage-launch

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

Launching a Claude Managed Agent requires assembling four ordered API payloads (environment, agent, session, kickoff) and executing them in sequence with correct ID chaining, all while keeping the Anthropic API key out of files, logs, and chat. Doing this by hand is error-prone and risks leaking credentials.

Core Features & Use Cases

  • Payload Generation: Convert a validated build-sheet.json into the four ordered CMA API payloads, with permission policies baked in (agent toolsets always_allow, MCP servers always_ask).
  • Resumable Launch Script: Write a launch.sh that creates environment → agent → session → kickoff in order, chaining IDs and resuming from the last created resource on re-run, reading $ANTHROPIC_API_KEY only at runtime.
  • Pre-Launch Validation: Run a validator that checks required fields, placeholder integrity, outcome rubrics, and scans all payloads and launch.sh for embedded API keys, failing the launch on any leak.
  • Use Case: After planning an agent with the interview phase, run the three scripts to produce payloads and launch.sh, validate them, then export your key and execute launch.sh to bring the agent online.

Quick Start

Ask the AI to generate the launch payloads and launch script from my build-sheet.json, validate them for key leaks, and walk me through launching with my own Anthropic API key.

Frequently Asked Questions about stage-launch

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I launch a Claude Managed Agent with curl?▼

Generate the four payloads with payload_generator.py, write launch.sh with launch_script_writer.py, validate with payload_validator.py, then export ANTHROPIC_API_KEY in your shell and run launch.sh. The script creates environment, agent, session, and kickoff in order, chaining IDs between steps.

How do I deploy an agent without exposing my Anthropic API key?▼

Keep the key only in your shell environment as $ANTHROPIC_API_KEY and never paste it into chat or files. The generated launch.sh reads the key at runtime, and payload_validator.py scans all payloads and the launch script for sk-ant- patterns, failing if any embedded key is found.

What happens if the launch script fails partway through?▼

Re-running launch.sh resumes from where it stopped. Each step writes its resource ID to a file (env.id, agent.id, session.id) and skips creation if that file already exists, so no duplicate resources are created.

Does the stage-launch skill make API calls itself?▼

No. All three Python tools are stdlib-only and make zero network calls. They only generate payloads, write the launch script, and validate files; the user personally runs launch.sh with their own key.

Why does payload validation fail before launch?▼

Validation fails on missing payloads, invalid JSON, missing agent name or model, a kickoff without a user.message event, an outcome with an empty rubric or max_iterations outside 1-20, or any detected API key leak. Fix the reported issue and re-run the validator.

Related Skills

Explore More Software Engineering Skills→
oakossoakoss
14

mcp-expert

Architect MCP ecosystems with server development, tool design, and transport configuration.

Official
Advanced
brianruggieribrianruggieri

eval-extract

Run blind extraction evaluations against fixtures and grade results with a rubric.

Community
Advanced
CodeAlive-AICodeAlive-AI
119

hooks-management

Create, update, and validate coding-agent hooks for Claude Code, Codex CLI, and OpenCode.

Official
Advanced
prapanchprapanch

iac-discipline

Enforce declarative Infrastructure as Code to prevent infrastructure drift.

Community
Intermediate