Phase 4 — Run Without You (the recurring loop) — Productivity

Phase 4 of building a Claude Managed Agent: turn a graded agent into a recurring scheduled deployment (POSIX-cron), an event-driven curl trigger, or confirmed on-demand use, then finalize the versioned roadmap. It validates cron schedules, builds deployment payloads, and writes NEXT-DIRECTIONS.md.

agent-launcher claude-code

📂 View skill on GitHub ↗ 🚀 How to use

✨ What it does

Validates 5-field cron expressions and IANA timezones with DST notes.
Builds POST /v1/deployments payloads, optionally nesting a self-grading outcome.
Prints BYOK curl commands to create and test the deployment.
Writes/refreshes NEXT-DIRECTIONS.md.
Enforces hard rules: manual test first, safety rails on by default, DST wall-clock awareness, ≤1,000 deployments/org.

🎯 When to use it

User says 'run it every morning', 'nightly', 'weekly', or 'automate this'.
User wants to put an agent on a recurring schedule or event-driven trigger.
The orchestrator routes to phase=run-without-you.
User wants to finalize the roadmap after a successful grade-iterate loop.

🚀 How to use

Trigger by saying 'run it every morning', 'put it on a schedule', 'nightly', 'weekly', 'automate this', or when the orchestrator routes phase=run-without-you. Inputs: a build-sheet JSON, agent ID, environment ID, cron expression, timezone, and optional --nest-outcome flag. Example prompts:

Run it every morning at 9 AM Berlin time.
Set up a nightly deployment with self-grading for my agent.

📄 Output: A validated deployment payload JSON, a NEXT-DIRECTIONS.md file, and printed curl commands for creating and testing the deployment.

📦 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/run-without-you .claude/skills/run-without-you

Skill source: agent-launcher/skills/run-without-you/SKILL.md

⚠️ Good to know

No tool makes API calls; the deployment is created via BYOK curl, and you must test with a manual run before trusting the schedule.

❓ FAQ

What does --nest-outcome do?

It includes the rubric in the deployment payload so each firing self-grades, nesting the grade→iterate loop inside every recurring run.

Why should I avoid 02:00–03:00 in DST zones?

Because wall-clock semantics mean spring-forward times are skipped and fall-back times fire twice, so exactly-once firing is not guaranteed during that hour.

What is the hard rule about testing?

Never commit a schedule you haven't fired once with a manual run; test with a manual run first and read the verdict before leaving the cron in place.

🤖 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.