AgentSkillsCN

write-planning-doc

为各类项目与计划制定结构化的规划文档,将复杂的工作分解为多个阶段,并明确各阶段的行动步骤与验证节点。

SKILL.md
--- frontmatter
name: write-planning-doc
description: "Creates structured planning documents for projects and initiatives, breaking down complex work into stages with clear actions and validation points."
last_updated: 2026-01-10
tools_required: [Write, Read, Bash]
agent_type: main_agent

Planning Document Guide

Guide for writing planning/project management .md files for any type of project or initiative.

[GOAL]

Create structured planning documents that break complex projects into stages with clear actions and validation points.

[CONTEXT]

Planning docs are for thinking through & documenting decisions, breaking down complex projects into multiple stages, and tracking progress.

Store planning docs in Cockpit/memory/topics/planning/ or the relevant project folder.

[PROCESS]

Starting a Planning Doc

  • Run date +"%y%m%d" to get the current date for naming
  • Ask the user questions about requirements to clarify key decisions
  • Use [[sounding-board-mode]] for discovery conversations

File Naming

Format: yyMMdd[letter]_description_in_lowercase.md

  • Date prefix: yyMMdd (e.g., 260110 for 10 Jan 2026)
  • Letter: auto-increment (a, b, c...) for same-day docs
  • Example: 260110a_athena_blue_launch_plan.md

Document Structure

markdown
# [Project Name] Planning Doc

## Goal, Context
- Clear problem/goal statement(s) at top
- Enough context to pick up where we left off

## References
- Relevant background docs, files, links
- 1-sentence summary for each
- Most important at top

## Principles, Key Decisions
- Specific approaches agreed with user
- Update as new information arrives

## Stages & Actions

### Stage: [Description]
- [ ] Top-level action
  - [ ] Sub-action
    - Notes with extra context

### ✅ Completed Stage
- ✅ Completed action
  - 📔 Notes on discoveries

## Appendix
- Research findings
- Alternative approaches considered
- Supporting materials

Stage Design Principles

  • Start with simple working v1, layer complexity gradually
  • End each stage with validated outputs
  • Don't number stages (easier to reorder)
  • Use [ ] and [x] checkboxes for todo/done
  • Surface potential risks early
  • Frontload business value

Validation Points

At end of stages, add appropriate checks:

  • Stakeholder review / sign-off
  • Quality review
  • Data validation
  • User testing
  • Stop & review with user

Final Actions

  • Final validation (stakeholder sign-off, QA)
  • Cleanup and documentation
  • Move to archive when complete

[IMPORTANT]

  • Keep concise but capture all decisions
  • Update regularly as work progresses
  • Stop immediately if surprises/issues - discuss first
  • Use subagents for encapsulated research tasks
  • Add web search for research, best practices, examples