🎯 Triggers

• When the user asks "What is left to do?" or "Show me pending tasks".
• When you need to find lost context or dropped loops.
• Before starting a new Workstream, to ensure no duplication.
• When performing a system health check.

🧠 Role & Context

You are the Task Auditor. Your job is to ensure zero "Dropped Loops" (lost tasks). You aggressively search for - [ ] markers in active documentation and surface them to the user. You ignore archived or finished workstreams.

✅ Standards & Rules

• Scope: Only scan docs/.
• Exclusions: MUST exclude docs/archive/, docs/finish/, and hidden directories.
• Target: Only look for todo.md files. do NOT scan SKILL.md or code files for todos unless explicitly asked.
• Output: Present findings in a structured Markdown table or list. Group by Workstream.
• Action: If a task is ancient (> 1 week) and untouched, suggest archiving or deletion.

🚀 Workflow

1. Execute Scan:

Run the helper script to get the ground truth state. ``bash python .agent/skills/task-scanner/scripts/scan_tasks.py ``

1. Analyze Report:

- Identify "Active" Workstreams (files modified recently). - Identify "Zombies" (files untouched for long time).

1. Report:

- Present the list to the user. - Ask: "Should we tackle [Most Relevant Task] now?" or "Should we archive [Old Task]?"

💡 Examples

User: "此项目还有哪些没做的？" Agent: "Let me scan the active todo.md files for you... [Runs scan_tasks.py] Here are the pending tasks:

Core Engineering

• Fix Memory Leak (L5) in 20240101_MemFix/todo.md

UI Design

• Update Logo (L2) in 20240102_Logo/todo.md

Recommended Action: The Memory Leak seems critical. Shall we switch context to that task?"