AgentSkillsCN

looper-live-test

通过创建一个临时项目（PROJECT.md + to-do.json），运行 Looper-Go 一次迭代，观察输出并记录结果，从而开展实时的 Looper-Go 冒烟测试。此方法适用于验证 Looper-Go 的行为，或用于演示实时运行过程。

SKILL.md
--- frontmatter
name: looper-live-test
description: Run a live Looper-Go smoke test by creating a temporary project (PROJECT.md + to-do.json), executing looper for one iteration, observing output, and reporting results. Use when verifying looper-go behavior or demonstrating a live run.

Looper-Go Live Test

Overview

Run a real looper-go iteration in a fresh temp directory to validate task selection and summary application, then report what happened.

Workflow

  1. •Locate the looper binary. Prefer looper on PATH; otherwise set LOOPER_BIN to a specific path (for example ~/looper-go/bin/looper or go run github.com/nibzard/looper-go/cmd/looper).
  2. •Run the helper script: ~/.codex/skills/looper-live-test/scripts/run-live-test.sh
  3. •Watch the output until it finishes; note the Task: line and Summary: line.
  4. •Inspect the temp project directory printed by the script; check to-do.json and any created files.
  5. •Report: selected task line, summary line, files created, and any errors or unexpected behavior. Mention the temp dir path.

Manual fallback (only if the script cannot run)

  1. •Create a temp directory and PROJECT.md.
  2. •Create a to-do.json with tasks T10 and T2 (both priority 1).
  3. •Run: MAX_ITERATIONS=1 looper run
  4. •Observe output, then inspect to-do.json and created files.

Looper-Go Specific Notes

  • •Looper-Go uses looper run as the main command (not looper.sh to-do.json)
  • •Configuration is loaded from .looper/config.json, environment, or CLI flags
  • •Logs are stored in looper-logs/runs/<run-id>/events.jsonl
  • •Use looper doctor to check dependencies and configuration