Claude Loop

Autonomous orchestrated loop that works through a prd.json story-by-story using a fresh Task subagent per story. Use this skill whenever the user asks to "run the loop", "claude loop", "start the loop", "kick off the loop", "work through the prd", "process prd.json", "implement all the stories", "iterate through the stories", "go through the user stories autonomously", or anything that suggests autonomously executing a list of user stories from a PRD file. Also trigger when the user references a prd.json file as their task list and wants implementation done across multiple iterations, or asks

49 stars 7 forksAI, LLMs & Agents

Install

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add aronprins/claude-loop --skill claude-loop
Manual install
git clone --depth 1 https://github.com/aronprins/claude-loop.git /tmp/claude-loop
mkdir -p ~/.claude/skills
cp -r /tmp/claude-loop/. ~/.claude/skills/claude-loop

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# Claude Loop

You are the **orchestrator** for an autonomous coding workflow. You do not implement stories yourself. You spawn a fresh **subagent** (via the `Task` tool) for each user story and coordinate their work until the PRD is complete.

You are the loop. There is no external script driving iterations — every iteration happens inside this single Claude Code session via `Task`.

## Two Modes

- **Sequential (default).** One subagent per story, one at a time, in priority order. Simplest and safest; stories may freely overlap on files. This is everything described up to **Parallel Mode** below — use it unless the user explicitly asks for parallel execution.
- **Parallel / wave (opt-in).** When the user asks to run the loop "in parallel", "in tandem", "as waves", or to "fan out", group stories into dependency **waves** and run each wave's independent stories concurrently — each in its own git worktree — with a merge+verify barrier between waves. This requires stories to declare a `dependsOn` array (see **Parallel Mode**). Everything in the sequential sections (preflight, stack detection, archive, completion signal) still applies; parallel mode only changes *how stories execute*.

---

## Where Things Live

The loop uses three distinct locations. Keep them straight:

| Location | What it is | Who writes to it |
|---|---|---|
| `.claude/skills/claude-loop/` | This skill's code (SKILL.md, subagent-prompt.md, etc.) | Nobody during a run — installed once |
Read the full skill on GitHub

Details

Skill name
claude-loop
License
MIT
Repo updated
22 June 2026

FAQ

What does the Claude Loop skill do?

Autonomous orchestrated loop that works through a prd.json story-by-story using a fresh Task subagent per story. Use this skill whenever the user asks to "run the loop", "claude loop", "start the loop", "kick off the loop", "work through the prd", "process prd.json", "implement all the stories", "iterate through the stories", "go through the user stories autonomously", or anything that suggests autonomously executing a list of user stories from a PRD file. Also trigger when the user references a prd.json file as their task list and wants implementation done across multiple iterations, or asks

How do I install the Claude Loop skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add aronprins/claude-loop --skill claude-loop. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Claude Loop popular?

It lives in aronprins/claude-loop, which has 49 GitHub stars and 7 forks.

Related skills

Connect Hubspot Agent CLI

81

TomGranot/hubspot-admin-skills

Install and connect HubSpot's Agent CLI so Claude Code can run bulk and administrative HubSpot operations directly from the terminal: JSONL bulk writes, deletes, property and pipeline management, and workflow CRUD, with dry runs. Covers user versus service key mode, safe-use rules, and how the CLI fits beside MCP and this repo's scripts.

AI, LLMs & AgentsIn 1 list19

Build Lead Scoring

81

TomGranot/hubspot-admin-skills

Create a comprehensive lead scoring model with separate Fit and Engagement scores using HubSpot's new Lead Scoring tool. Replaces the deprecated HubSpot Score property.

AI, LLMs & AgentsIn 1 list19

Connect Hubspot MCP

81

TomGranot/hubspot-admin-skills

Connect Claude to HubSpot's official remote MCP server for conversational CRM lookups, spot-checks, and small confirmed edits. Covers the MCP connector setup with OAuth and PKCE, what the server's tools can and cannot do, and when to use MCP versus the Agent CLI versus this repo's scripts.

AI, LLMs & AgentsIn 1 list19

Sector Performance Snapshot

132

OctagonAI/skills

Retrieve a snapshot of market sector performance using Octagon MCP. Use when analyzing sector-wide metrics including revenue, EBITDA, net income, market cap, and enterprise value for companies within a specific sector and exchange.

AI, LLMs & Agents14

Disk Cleaner

49

open-agent-power/oh-my-disk-cleaner

Cross-platform disk space management toolkit with intelligent optimization. REQUIREMENTS: Python 3.7+. UNIVERSAL COMPATIBILITY: Works with ALL AI IDEs (Cursor, Windsurf, Continue, Aider, Claude Code, etc.). PLATFORM-INDEPENDENT: Works at any location - global, project, or user level. SELF-CONTAINED: No pip install needed, includes intelligent bootstrap. KEY FEATURES: (1) PROGRESSIVE SCANNING: Quick sample (1s) + Progressive mode for large disks, (2) INTELLIGENT BOOTSTRAP: Auto-detection of skill location and auto-import of modules, (3) CROSS-PLATFORM ENCODING: Safe emoji/Unicode handling on al

AI, LLMs & Agents

Sector Pe Ratios

132

OctagonAI/skills

Retrieve sector P/E ratios using Octagon MCP. Use when comparing company valuations to sector benchmarks, analyzing sector valuations across exchanges, and understanding market-wide valuation trends.

AI, LLMs & Agents14

Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.