← Back to Skills Marketplace
austindixson

Agent Loops

by austindixson · GitHub ↗ · v2.1.0
cross-platform ⚠ suspicious
386
Downloads
0
Stars
2
Active Installs
2
Versions
Install in OpenClaw
/install agent-loops
Description
Multi-agent workflow orchestrator. Use when the user asks to build, create, make, ship, develop, or launch any software (apps, webapps, websites, mobile apps...
README (SKILL.md)

Agent Loops

Prebuilt multi-agent workflows that chain sequential and parallel steps, with real output passing between agents.

Description

Agent Loops orchestrates multi-step agent pipelines. Each workflow defines a sequence of steps; each step runs via claude -p with a role-specific system prompt and agent-swarm model routing. Outputs chain between steps so each agent builds on the previous one's work.

Use when the user wants to:

  • Build, create, or ship anything — apps, webapps, websites, mobile apps (iPhone/Android), desktop apps, APIs, CLIs, bots, dashboards, landing pages, SaaS products, MVPs, prototypes, plugins, extensions, microservices
  • Fix, debug, troubleshoot, or diagnose bugs, errors, crashes, or failing tests
  • Review, audit, or inspect code for bugs, security vulnerabilities, or quality issues
  • Research, investigate, compare, or write reports on any topic
  • Refactor, restructure, clean up, optimize, or modernize code
  • Test, review, and publish a skill to ClawHub

Installation

clawhub install agent-loops

Or clone into your skills directory:

git clone https://github.com/OpenClaw/agent-loops.git workspace/skills/agent-loops

Requires PyYAML for YAML workflows:

pip install pyyaml

Usage

Workflow selection — match the user's intent to a workflow:

User says something like... Workflow
"build me an app", "create a webapp", "make a website", "ship this feature", "develop a mobile app", "launch a SaaS", "make an iPhone app", "build an Android app", "create a desktop app", "make a CLI tool", "build an API", "create a bot", "make a dashboard", "build a landing page", "create an MVP", "prototype X", "add dark mode", "implement Y", "build a plugin", "make an extension", "create a service", "spin up a microservice", "scaffold a project" ship_feature
"fix this bug", "debug X", "why is Y broken", "troubleshoot this error", "diagnose the crash", "this isn't working", "something's wrong with X", "getting an error when", "it crashes on", "the build is failing", "tests are broken", "patch this issue", "hotfix for X" bug_fix
"review this code", "audit the codebase", "check for security issues", "inspect this PR", "find bugs in X", "analyze this module", "is this code safe", "check for vulnerabilities", "look over my changes", "do a security review", "scan for issues", "evaluate code quality" code_review
"research X", "compare A vs B", "write a report on", "investigate Y", "explore options for", "what are the best practices for", "study the landscape of", "deep dive into", "summarize the state of", "pros and cons of", "analyze the market for", "write up findings on" research_report
"refactor X", "clean up this code", "restructure Y", "reorganize the codebase", "optimize this module", "modernize the architecture", "reduce tech debt", "simplify this", "extract a service", "decouple X from Y", "improve code structure", "make this more maintainable" refactor
"publish this skill", "push to ClawHub", "release this skill", "deploy to ClawHub", "ship this skill", "get this skill ready for publish" skill_publish

Command pattern:

python3 workspace/skills/agent-loops/scripts/run_workflow.py \x3Cworkflow> "\x3Cuser's request>" --apply

Pass the user's natural language request as the input. The workflow handles scoping, delegation, and output chaining automatically.

Examples

Example 1: Ship a feature Scenario: You want to scope, implement, and document a new feature. Action: python3 workspace/skills/agent-loops/scripts/run_workflow.py ship_feature "Add dark mode toggle to settings" --apply Outcome: PM scopes tasks, Dev implements with tests, Editor writes docs and changelog.

Example 2: Fix a bug Scenario: A bug needs diagnosis, a fix, and regression tests. Action: python3 workspace/skills/agent-loops/scripts/run_workflow.py bug_fix "Login page crashes on empty password" --apply Outcome: Dev diagnoses root cause, Dev implements fix, Tester writes regression test, Editor documents the change.

Example 3: Parallel code review Scenario: You want a code review and security audit run simultaneously. Action: python3 workspace/skills/agent-loops/scripts/run_workflow.py code_review "Review the auth module" --apply Outcome: Reviewer and Security auditor run in parallel, then Editor synthesizes a unified summary.

Example 4: Override model Scenario: You want all steps to use a specific model. Action: python3 workspace/skills/agent-loops/scripts/run_workflow.py research_report "Compare REST vs GraphQL" --apply --model sonnet Outcome: All steps run with the specified model instead of agent-swarm routing.

Commands

python3 workspace/skills/agent-loops/scripts/run_workflow.py \x3Cworkflow> "\x3Cinput>"              # Dry-run a workflow
python3 workspace/skills/agent-loops/scripts/run_workflow.py \x3Cworkflow> "\x3Cinput>" --apply       # Run agents for real
python3 workspace/skills/agent-loops/scripts/run_workflow.py --list                             # List available workflows
python3 workspace/skills/agent-loops/scripts/run_workflow.py --list --json                      # List workflows as JSON
python3 workspace/skills/agent-loops/scripts/run_workflow.py \x3Cworkflow> "\x3Cinput>" --apply --json # Output results as JSON
python3 workspace/skills/agent-loops/scripts/run_workflow.py \x3Cworkflow> "\x3Cinput>" --model sonnet # Override model for all steps
python3 workspace/skills/agent-loops/scripts/run_workflow.py \x3Cworkflow> "\x3Cinput>" --timeout 900  # Set per-step timeout (seconds)
python3 workspace/skills/agent-loops/scripts/run_workflow.py \x3Cworkflow> "\x3Cinput>" -v             # Verbose output
  • \x3Cworkflow> — Workflow id: ship_feature, bug_fix, code_review, research_report, refactor, skill_publish
  • --apply — Actually spawn agents (default is dry-run)
  • --list — List all available workflows
  • --json — Output results as JSON for programmatic use
  • --model — Override agent-swarm routing with a specific model
  • --timeout — Per-step timeout in seconds (default: 600)
  • -v — Show full task text per step
Usage Guidance
What to check before installing/using: - Expect the script to call the 'claude' CLI (claude -p). If you don't have or want that CLI, do not run --apply. The CLI will send task text to its backend (so anything you pass may be transmitted externally). - The metadata omits this requirement; manually verify you have 'claude' in PATH and understand its authentication and privacy model. - The script reads OPENCLAW_HOME (defaults to ~/.openclaw) and will write run records to OPENCLAW_HOME/workspace/skills/agent-loops/runs — inspect that directory and its contents; avoid passing secrets in requests if you don't want them stored or transmitted. - If an agent-swarm router script exists at OPENCLAW_HOME/workspace/skills/agent-swarm/scripts/router.py, this skill will execute it (subprocess). Audit that local router script before running to avoid executing untrusted local code. - If you want to be cautious: run the script in dry-run mode (omit --apply) to see planned steps, review run_workflow.py and workflows, and run in an isolated environment or sandbox before giving it real project data. - The inconsistencies (missing declared binary/env requirements) likely indicate sloppy metadata, not necessarily malicious intent, but you should inspect and confirm the runtime behavior matches your security/privacy needs.
Capability Analysis
Type: OpenClaw Skill Name: agent-loops Version: 2.1.0 The agent-loops skill is a multi-agent orchestrator designed to chain sequential and parallel LLM tasks using the Claude CLI. The core logic in `scripts/run_workflow.py` safely handles task execution via subprocess lists (avoiding shell injection) and includes features for run persistence and auditing. The workflow definitions in the `workflows/` directory and the instructions in `SKILL.md` are consistent with the stated purpose of software development, research, and code review, with no evidence of malicious intent, data exfiltration, or unauthorized persistence.
Capability Assessment
⚠ Purpose & Capability
The skill claims to orchestrate agent workflows and the included script implements that by invoking a 'claude' CLI and an optional agent-swarm router. However, registry metadata declares no required binaries or env vars even though the README and run_workflow.py require the 'claude' CLI in PATH and read OPENCLAW_HOME. The omission of these runtime dependencies in metadata is an incoherence the user should notice.
ℹ Instruction Scope
SKILL.md instructs running the included Python script which will spawn subprocesses (claude CLI) to execute tasks, chain outputs, and (per README) persist runs to a runs/ directory. The instructions do not attempt to read unrelated system files, but they do cause data to be sent to whatever backend the 'claude' CLI uses and they will save run output locally.
✓ Install Mechanism
No install spec is provided (instruction-plus-code only). This reduces supply-chain risk from remote downloads. The repository does include requirements.txt recommending PyYAML; SKILL.md suggests pip install pyyaml.
⚠ Credentials
The skill declares no required environment variables or credentials but the code reads OPENCLAW_HOME (defaulting to ~/.openclaw) and passes the full environment to subprocesses. It relies implicitly on a configured 'claude' CLI (which may itself require credentials). The lack of declared env/binary requirements is disproportionate and hides important runtime assumptions.
ℹ Persistence & Privilege
The script persists live-run data to OPENCLAW_HOME/.../runs (per README and RUNS_DIR in code). always:false and it does not request elevated privileges or modify other skills. Persisting user inputs/outputs locally is normal for tooling but should be noted since sensitive inputs will be saved by default.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install agent-loops
  3. After installation, invoke the skill by name or use /agent-loops
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.1.0
- Expanded and clarified skill description to better cover all major supported workflow types and user requests. - Added a workflow intent table to help match common user requests to the appropriate workflow for faster onboarding. - Updated examples and usage guidance for stronger real-world applicability. - No changes to commands or arguments.
v2.0.0
Major update: Agent Loops 2.0 introduces robust multi-agent workflow orchestration, supporting sequential and parallel pipelines with real output passing. - Prebuilt workflows for complex tasks (feature shipping, bugfix, code review, research, skill publishing) - Seamless chaining of agent steps with support for both sequential and parallel execution - Simple command-line interface for dry-run, execution, model override, output format, and timeout controls - Custom workflows enabled via YAML files - PyYAML dependency for workflow configuration - Expanded examples and full usage instructions for orchestrating agent-swarm pipelines
Metadata
Slug agent-loops
Version 2.1.0
License —
All-time Installs 2
Active Installs 2
Total Versions 2
Frequently Asked Questions

What is Agent Loops?

Multi-agent workflow orchestrator. Use when the user asks to build, create, make, ship, develop, or launch any software (apps, webapps, websites, mobile apps... It is an AI Agent Skill for Claude Code / OpenClaw, with 386 downloads so far.

How do I install Agent Loops?

Run "/install agent-loops" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Agent Loops free?

Yes, Agent Loops is completely free (open-source). You can download, install and use it at no cost.

Which platforms does Agent Loops support?

Agent Loops is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Agent Loops?

It is built and maintained by austindixson (@austindixson); the current version is v2.1.0.

More Skills
self-improving agent
Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
⬇ 463283 · by pskoett
Skill Vetter
Security-first skill vetting for AI agents. Use before installing any skill from ClawdHub, GitHub, or other sources. Checks for red flags, permission scope, and suspicious patterns.
⬇ 259410 · by spclaudehome
Polymarket
Query Polymarket prediction markets. Check odds, find trending markets, search events, track price movements.
⬇ 232503 · by joelchance
Self-Improving + Proactive Agent
Self-reflection + Self-criticism + Self-learning + Self-organizing memory. Agent evaluates its own work, catches mistakes, and improves permanently. Use when...
⬇ 200423 · by ivangdavila
Github
Interact with GitHub using the `gh` CLI. Use `gh issue`, `gh pr`, `gh run`, and `gh api` for issues, PRs, CI runs, and advanced queries.
⬇ 191115 · by steipete
ontology
Typed knowledge graph for structured agent memory and composable skills. Use when creating/querying entities (Person, Project, Task, Event, Document), linkin...
⬇ 190156 · by oswalpalash

💬 Comments