copal-engineering-loop
Use when implementing features with structured TDD workflow, task tracking via todo.json, and engineering best practices.
📖 中文导读:Use when implementing features with structured TDD workflow, task tracking via todo.json, and engineering best practices. 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。
name: Copal Engineering Loop description: Use when implementing features with structured TDD workflow, task tracking via todo.json, and engineering best practices.
Copal Engineering Loop Skill
This skill integrates with the Copal CLI to provide structured, state-driven development workflows.
When to Use This Skill
Activate this skill when:
- User asks to implement a feature or fix a bug
- User wants structured development workflow (Plan → Research → Work → Review)
- Project has
.copal/directory withtodo.json
Available Commands
Task Management
# Get and start the next pending task
copal next [--worktree]
# Mark a task as completed
copal done <task_id>
# Show project status, task summary, and recent session history
copal status
# Validate project state before starting work
copal validate --pre-task
# Search project memory
copal memory search --query "search term"
Workflow
1. Session Initialization (Get Bearings)
- Run
copal statusto see active tasks. - Run
copal nextto get context and see recent session history. - If you need more context, use
copal memory search. - Run
copal validate --pre-taskto ensure:- Git working directory is clean
- Baseline tests are passing
2. Execution Loop
- Run
copal next(orcopal next --worktreefor isolation) to claim a task. - Worktree: If using worktree,
cdinto the new directory. - TDD: Write tests -> Fail -> Implement -> Pass.
- Commit: Make atomic commits for the task.
3. Completion & Handover
- Run
copal done <id>. This will:- Mark task as done in
todo.json - Save a session summary to memory automatically
- Mark task as done in
- If using worktree, merge changes back to main branch.
Task States
| Status | Meaning |
|---|---|
todo |
Not started |
in_progress |
Currently being worked on |
done |
Completed |
blocked |
Cannot proceed |
Best Practices
- Trust the Harness: Rely on
copal validateand memory chains. - One Feature, One Branch: Use
copal next --worktreefor complex tasks. - Validation Gates: Never start a task on a dirty or broken state.
- Self-Correction: If
copal donefails, fix the issue and retry.
File Locations
- Task list:
.copal/artifacts/todo.json - Memories:
.copal/memory.sqlite(managed via CLI)