AgentSkillsCN

update-changelog

在实施过程中，将阶段完成详情或最终总结追加至changelog.md。

SKILL.md
--- frontmatter
name: update-changelog
description: Append phase completion details or final summary to changelog.md during implementation

Update Changelog

Append phase tracking or final summary to thoughts/NNNN-description/changelog.md.

When to Use

During Implementation (after each phase):

  • •Document what was actually done vs planned
  • •Note deviations and reasons
  • •List files changed
  • •Capture discoveries

At Completion (after all phases):

  • •Append FINAL SUMMARY section
  • •Summarize key deviations
  • •Document lessons learned
  • •Note technical debt and follow-up work

Phase Update Format

Use template from templates/changelog-document.md Phase section:

markdown
## Phase {N}: {Phase Name}
**Completed**: {ISO_DATE}
**Status**: ✅ Complete

### What We Did
- Created slug determination skill
- Added interactive prompting with suggestions

### Deviations from Plan
- **Planned**: Create bash script only
- **Actually**: Created both skill and script
- **Reason**: Skill needed for integration with commands

### Files Changed
- `skills/determine-feature-slug/SKILL.md` - New skill file
- `scripts/next-feature-slug.sh` - Helper script

### Discoveries
- Bash parameter expansion handles leading zeros: `10#$NUM`
- Kebab-case validation regex: `^[a-z0-9-]+$`

Final Summary Format

Use template from templates/changelog-document.md FINAL SUMMARY section:

markdown
## 🎯 FINAL SUMMARY
**Completion Date**: 2026-02-05
**Overall Status**: ✅ Complete

### What Was Built
Full plugin restructuring with feature directories, changelog tracking,
milestone grouping, and agent orchestration.

### Key Deviations
1. Added interactive slug prompting (not in original plan)
2. Kept thoughts/notes/ unchanged instead of moving

### Impact on Original Plan
- Plan estimated 11 phases, completed 11
- Added 2 skills not in plan (prompting helpers)
- Simplified 3 phases by combining related changes

[... rest of sections ...]

Important Guidelines

  • •Read plan first to understand what was intended
  • •Compare actual vs planned - be specific about deviations
  • •Capture learnings - what would you do differently?
  • •Note technical debt - what was compromised for speed?
  • •Append only - never modify previous phase entries
  • •Be concise - focus on what matters, skip obvious details

Auto-Correction Loop

The changelog enables auto-correction:

  1. •Before Phase N: Read both plan.md AND changelog.md
  2. •Plan shows original intent
  3. •Changelog shows actual state after Phase N-1
  4. •Adapt Phase N approach based on actual state
  5. •After Phase N: Document new deviations

Example:

code
Phase 3 needs to integrate with auth system
- plan.md says: "Use middleware pattern from Phase 1"
- changelog.md says: "Phase 1 used decorator pattern instead"
- Agent adapts: Uses decorator pattern, not middleware ✅