skills.rest
freedom909
freedom909COMMUNITY
Explore all skills by this author →
GitHub

autonomous-loops

Implements autonomous Claude Code loop architectures from sequential pipelines to multi-agent DAG orchestration.

Updated Mar 18, 2026
Search Tags:#automation#ci/cd#claude code#workflow orchestration#autonomous agents#multi-agent#dag
One-click install
npx skills add https://github.com/freedom909/real-estate-saas --skill autonomous-loops-freedom909
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: autonomous-loops
Source: https://github.com/freedom909/real-estate-saas/tree/main/.trae/skills/autonomous-loops
Command: npx skills add https://github.com/freedom909/real-estate-saas --skill autonomous-loops-freedom909

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Running Claude Code autonomously over many iterations requires choosing the right loop architecture, bridging context between isolated runs, and coordinating parallel agents without merge conflicts or quality degradation. ## Core Features & Use Cases - Loop Pattern Spectrum: Six patterns from simple claude -p sequential pipelines and the NanoClaw REPL to the Infinite Agentic Loop, Continuous Claude PR loop, the De-Sloppify cleanup pass, and RFC-driven DAG orchestration. - Cross-Iteration Context: Uses SHARED_TASK_NOTES.md and filesystem state to bridge context between independent non-interactive invocations. - Quality Gates & Merge Coordination: Tiered pipelines (research, plan, implement, test, review) with separate context windows per stage and an agent-driven merge queue with eviction recovery. - Use Case: You have a written RFC for a large feature. Decompose it into a dependency DAG of work units, run each through a tiered quality pipeline in isolated worktrees, and land them via a merge queue that captures conflict context for re-runs. ## Quick Start Set up an autonomous Claude Code loop for my project by choosing the right pattern from the autonomous-loops skill and generating the pipeline script.

Frequently Asked Questions about autonomous-loops

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

FAQPage Schema
How do I run Claude Code autonomously in a loop?▼

Chain non-interactive claude -p calls in a shell script, one focused step per call, with set -e to stop on failure. For continuous iteration, use the Continuous Claude PR loop which creates branches, opens PRs, waits for CI, and merges automatically.

What is the difference between a sequential pipeline and the Continuous Claude loop?▼

A sequential pipeline runs fixed claude -p steps in order with fresh context each step, suiting scripted daily workflows. Continuous Claude iterates indefinitely with PR creation, CI polling, failure auto-fix, and limits like max-runs, max-cost, or max-duration.

How do I persist context between Claude Code loop iterations?▼

Use a SHARED_TASK_NOTES.md file that Claude reads at iteration start and updates at iteration end. This bridges the context gap because each claude -p invocation starts with a fresh context window.

When should I use RFC-driven DAG orchestration instead of simpler loops?▼

Use it when a written RFC decomposes into multiple interdependent work units needing parallel implementation with merge coordination. For single focused changes or quick iteration, a sequential pipeline or NanoClaw REPL is sufficient.

Why add a separate de-sloppify cleanup pass instead of negative instructions?▼

Negative instructions like "don't test type systems" make the model hesitant about all testing and degrade quality unpredictably. A separate cleanup pass removes redundant tests and over-defensive checks while keeping business logic tests intact.

How do I prevent infinite loops from running forever?▼

Always set an exit condition: max-runs for iteration count, max-cost for spending, max-duration for time, or a completion signal phrase. Continuous Claude stops after a configurable threshold of consecutive completion signals.

Related Skills

Explore More Software Engineering Skills→
ximingximing

ralph

Convert Product Requirement Documents into structured prd.json for Ralph.

Community
Intermediate
bravrosbravros
3

finish

Merge approved pull requests, close plans, and route homolog-to-main decisions.

Official
Advanced
FaisalAlqarniFaisalAlqarni
5

verification-loop

Run build, type, lint, test, security, and diff checks before a PR.

Community
Intermediate
granietgraniet
247

claude-code

Delegate coding tasks to the Claude Code CLI for automation.

Community
Advanced