AgentSkillsCN

task-parsing

将自然语言解析为结构化的任务,明确任务优先级、截止日期、所属项目以及标签信息。适用于新增任务、处理待办事项,或当用户描述待办工作时使用。

SKILL.md
--- frontmatter
name: task-parsing
description: Parses natural language into structured tasks with priority, due date, project, and tags. Use when adding tasks, processing inbox, or when user describes work to be done.

Task Parsing

Convert natural language into structured task format.

Task Format

code
- [ ] Task description @project #tag !priority ~estimate

Parsing Rules

Priority Detection

Input PatternPriority
"urgent", "ASAP", "critical", "blocking"!high
"important", "should", "need to"!medium
"when possible", "eventually", "nice to have"!low

Time Estimation

Input PatternEstimate
"quick", "small", "minor"~15m
"short", "simple"~30m
"medium", "regular"~1h
"large", "significant"~2h
"big", "complex", "deep work"~4h

Due Date Extraction

Input PatternAction
"today", "EOD"Add to today.md
"tomorrow"Add to today.md with note
"this week", "by Friday"Add to backlog, Do First
"next week", "soon"Add to backlog, Schedule
"someday", "eventually"Add to someday.md

Project Detection

  • Look for explicit @project-name mentions
  • Infer from context: "for the API project" → @api
  • Work-related defaults to @work
  • Personal items get #personal tag

Tag Extraction

Common tags to apply:

  • #code - Programming tasks
  • #docs - Documentation
  • #meeting - Meetings and calls
  • #research - Research and learning
  • #bug - Bug fixes
  • #feature - New features
  • #review - Code/document reviews

Examples

Input: "Review John's PR for the auth feature, it's blocking the release" Output: - [ ] Review John's PR for auth feature @work #review #code !high ~30m

Input: "Eventually learn Kubernetes" Output: - [ ] Learn Kubernetes #learning !low ~4h → someday.md

Input: "Need to fix the login bug today" Output: - [ ] Fix login bug @work #bug !high ~1h → today.md

Processing Steps

  1. Extract the core action (verb + object)
  2. Detect priority from urgency words
  3. Estimate time from complexity hints
  4. Extract or infer project
  5. Add relevant tags
  6. Determine destination file
  7. Format as markdown checkbox