ccw-loop_skill

This skill orchestrates stateless iterative development loops with file-based state tracking, enabling auto or manual develop, debug, and validate phases.
  • Python

1.2k

GitHub Stars

2

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill catlog22/claude-code-workflow --skill ccw-loop

  • README.md4.3 KB
  • SKILL.md12.0 KB

Overview

This skill implements a stateless iterative development loop for file-based progress tracking and resumable workflows. It supports develop, debug, and validate phases and can run interactively or in automatic cycles. The workflow is driven by a unified state file and progress directory for transparent, resumable execution.

How this skill works

The skill creates or resumes a loop identified by a loop ID and stores master state in .workflow/.loop/{loopId}.json. It spawns a single orchestrator agent that uses spawn_agent, wait, send_input, and close_agent to execute multi-phase actions and write progress into .workflow/.loop/{loopId}.progress/. The agent checks control signals (running/paused/failed) before each action and updates state and progress files after each step.

When to use it

  • Start new feature, bugfix, or test-driven task that benefits from documented iterations.
  • Resume interrupted development work with preserved state and progress files.
  • Run automated cycles for standard develop->debug->validate flows.
  • Use interactive mode for complex tasks requiring manual choices.
  • Integrate with a dashboard or CLI for pause/resume/stop control.

Best practices

  • Keep project guidelines and tech context files in .workflow so the orchestrator can read them first.
  • Prefer single-agent deep interaction: use send_input to continue phases rather than spawning many agents.
  • Write meaningful progress entries (develop.md, debug.md, validate.md) so reviews are traceable.
  • Use --loop-id to resume instead of creating duplicate loops.
  • Set sensible max_iterations and add checkpoints to avoid runaway loops.

Example use cases

  • Implement a new endpoint: run develop to add code, debug to reproduce failures, validate with tests, and produce summary.md.
  • Triage and fix a production bug: resume the loop, reproduce in debug, apply fixes, and validate.
  • Automated CI task: trigger auto-loop to run develop->debug->validate for routine maintenance.
  • Collaborative sessions: share loop ID so multiple contributors can resume and inspect progress files.

FAQ

Pass --loop-id with the loop ID or call the skill with the existing loop ID. The skill will read .workflow/.loop/{loopId}.json and continue from the saved state.

What happens when the loop status is paused or failed?

The orchestrator checks the status before each action. If status is paused or failed the agent exits gracefully and awaits resume or restart commands; no progress is lost.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational
ccw-loop skill by catlog22/claude-code-workflow | VeilStrat