CommunityCoding & Developmentgithub.com

103kaz/claude-code-loops

Create .claude/verify.sh for the review-loop skill's verify step. Use when asked to set up or fix the verify loop.

What is claude-code-loops?

claude-code-loops is a Claude Code agent skill that create .claude/verify.sh for the review-loop skill's verify step. Use when asked to set up or fix the verify loop.

Works with✓Claude Code~Codex CLI~Cursor
npx skills add https://github.com/103kaz/claude-code-loops/tree/HEAD/skills/init-verify

Installed? Explore more Coding & Development skills: steipete/bluebubbles, steipete/eightctl, steipete/blucli · View all 6 →

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

init-verify

Propose a fast .claude/verify.sh, show it, and write it only after the user agrees. Keep reads small: this skill should cost few tokens.

Steps

  1. List the project root once (names only). Read at most the build files that exist: package.json (the scripts block only), pyproject.toml, go.mod, Cargo.toml, Makefile. Do not read source files or CI logs.
  2. If .claude/verify.sh exists, read it and propose a diff instead of a rewrite.
  3. Pick only checks that finish in seconds: lint, type check, unit tests, using commands the project already defines. Leave slow suites out and say so.
  4. Draft the script (set -euo pipefail, cd "$(dirname "$0")/..", one line per check, no echo noise: a failure's output goes back to Claude, so quiet commands are cheaper). A non-zero exit means the verification failed.
  5. Show the script and which checks were left out, then ask. Write it only after a yes.
  6. Run it once and report the result and the time. If nothing can be verified quickly, say so; do not write a script that always passes.

Individual skills in this repo

This repo contains 2 individual skills — each has its own dedicated page.

Related Skills