AgentSkillsCN

Daily Prep

当用户说“/daily-prep”“每日备课”“晨间备课”“今天我该专注什么”“规划一天”“开启新的一天”“为今天做准备”,或希望获得关于今日工作重点的建议时,应使用此技能。它能结合能量与日历背景,为你带来一份充满干劲的晨间承诺。

SKILL.md
--- frontmatter
name: Daily Prep
description: This skill should be used when the user says "/daily-prep", "daily prep", "morning prep", "what should I focus on today", "plan my day", "start my day", "prep for today", or wants help deciding what to work on. Creates a morning commitment with energy and calendar context.
version: 0.1.0

Daily Prep

A morning ritual for deciding what to focus on today. The skill guides the user through energy check, calendar assessment, surfacing relevant tasks, and finalizing 1-3 commitments.

Philosophy

The value is not in filtering algorithms; it is in creating an evaluable contract with yourself. Morning commitment + evening reflection forms the bookend.

Research supports this approach:

  • Implementation intentions (specific commitments) improve goal attainment ~3x
  • Energy management outperforms pure time management
  • Zeigarnik effect: planning closes open loops even without completion
  • Decision fatigue: keep task selection simple and early

Morning Flow

1. Energy Check

To begin, use AskUserQuestion to collect the user's energy level:

code
Question: "How's your energy this morning?"
Options:
- Sharp: Ready for deep work
- Steady: Normal energy, can handle variety
- Low: Need easy wins or lighter tasks

2. Calendar Shape

Next, use AskUserQuestion for meeting density:

code
Question: "What does your calendar look like?"
Options:
- Clear: Large blocks of uninterrupted time
- Scattered: Some meetings but workable gaps
- Heavy: Meeting-dense, need quick tasks

3. Surfacing

To surface relevant work, reference the vault-task-management skill for task queries. Use Grep and Read tools to find recent captures.

Look for items in these categories:

Pressure - Tasks with external weight:

  • Fire tasks (f status in vault-task-management)
  • Tasks mentioned in recent daily notes
  • Tasks connected to people (names in task text)
  • Approaching deadlines

Slipping - Tasks going cold:

  • Tasks in files not modified for 3+ days
  • Incomplete tasks from old daily notes
  • Items mentioned multiple times without progress

Quick Wins - Small completable items:

  • Short task text
  • No dependencies mentioned
  • Similar to tasks completed quickly before

Present surfaced items grouped by category. Example:

code
Here's what I found in your vault:

**Pressure**
- [f] Review Roman's PR (mentioned in yesterday's note)
- TTS spec response to Ryan (people dependency)

**Slipping**
- Auth ADR from last week (not touched in 5 days)
- Test coverage for session manager (3 mentions, no progress)

**Quick Wins**
- Update CLAUDE.md with new pattern
- Close stale GitHub issue

4. Dialogue

Allow the user to respond with corrections, additions, or redirections:

  • "What about that other thing?"
  • "Actually, the ADR is blocked"
  • "Add reviewing the SDK migration"

Incorporate feedback and adjust the surfaced list. The user drives; the skill supports.

5. Commitment

To finalize, suggest 1-3 commitment items based on:

  • Energy level (Sharp = deep work OK, Low = quick wins)
  • Calendar shape (Heavy = only time-boxed items)
  • User's expressed priorities during dialogue

Present as a proposal, not a mandate:

code
Based on your steady energy and scattered calendar, I'd suggest:

1. Review Roman's PR with real feedback
2. Revisit auth ADR from last week
3. TTS spec response to Ryan

Does this feel right for today?

Use AskUserQuestion to confirm or allow the user to edit in their own words.

6. Save

Use the Write tool to save the daily prep file.

Path: {inboxPath}/daily-prep/YYYY-MM-DD.md

Where inboxPath is the vault's configured inbox directory (usually 00_Inbox).

Required format: The file MUST have YAML frontmatter followed by markdown body:

yaml
---
date: YYYY-MM-DD
energy: [user's response, lowercase: sharp/steady/low]
calendar: [user's response, lowercase: clear/scattered/heavy]
commitment:
  - text: "[first commitment]"
    assessment: null
  - text: "[second commitment]"
    assessment: null
---

Then add the markdown body:

markdown
# Daily Prep: YYYY-MM-DD

## Morning

**Energy**: [Title case]
**Calendar**: [Title case]

### What I'm Committing To

1. [First commitment]
2. [Second commitment]

For complete schema (evening closure, field definitions), see references/file-format.md.

Evening Integration

This skill creates prep files that /daily-debrief can read for evening closure. When running /daily-debrief:

  1. Check for today's prep file
  2. If exists, show the morning commitment
  3. Collect assessment for each item (Done / Partial / Blocked / Skipped)
  4. Record reflection and update the file

The skill does not implement evening closure directly; it prepares the data structure that debrief uses.

Tools Used

No custom scripts. Uses existing Claude Code tools:

  • AskUserQuestion - Structured input for energy, calendar, confirmation
  • Grep - Search vault for tasks, names, recent mentions
  • Glob - Find files by date pattern
  • Read - Read daily notes and task files
  • Write - Save the daily prep file

Reference vault-task-management skill for task query patterns.

Example Session

User: "/daily-prep"

Claude: "How's your energy this morning?" [AskUserQuestion: Sharp / Steady / Low]

User selects: Steady

Claude: "What does your calendar look like?" [AskUserQuestion: Clear / Scattered / Heavy]

User selects: Scattered

Claude: [Searches vault, presents surfaced items] "Here's what I found..."

User: "That looks right, but add the SDK migration review"

Claude: "Got it. Based on steady energy and scattered calendar:

  1. Review Roman's PR with real feedback
  2. Revisit auth ADR from last week
  3. SDK migration review

Ready to commit to these?"

User: "Yes"

Claude: [Writes prep file, confirms] "Done. Your commitment for today is saved. I'll check in during debrief."

Constraints

  • Suggest 1-3 items (research warns against over-planning)
  • Do not enforce a limit; trust user judgment
  • No calendar API integration in v1 (self-report only)
  • Pattern feedback is future scope (not implemented)