スキル一覧に戻る
savvy-web

design-prune

by savvy-web

0🍴 0📅 2026年1月25日
GitHubで見るManusで実行

SKILL.md


name: design-prune description: Remove historical cruft from design docs. Use when docs accumulate outdated content, after major refactorings, or when preparing for releases. allowed-tools: Read, Edit, Glob, Bash context: fork agent: design-doc-agent

Design Documentation Pruning

Removes historical cruft from design documentation by detecting and removing outdated content, update histories, and deprecated sections while preserving essential historical context.

Overview

This skill cleans up design documentation by identifying historical content patterns, calculating staleness scores, generating pruning plans, and applying safe removal strategies to keep docs focused on current state while preserving important context.

Quick Start

Basic pruning:

/design-prune effect-type-registry

Aggressive pruning:

/design-prune effect-type-registry --aggressive

Dry-run (preview only):

/design-prune effect-type-registry --dry-run

Parameters

Required

  • module: Module name to prune (or "all" for all modules)

Optional

  • doc: Specific document to prune (default: all docs in module)
  • aggressive: Remove all historical content (default: false, conservative mode)
  • dry-run: Preview changes without applying (default: false)

Workflow

High-level pruning process:

  1. Parse parameters to determine module, document, and pruning mode
  2. Load design.config.json to get module paths and pruning configuration
  3. Read design document and parse structure, sections, and content types
  4. Detect historical content using pattern matching for historical markers
  5. Calculate staleness score based on age, markers, redundancy, and cross-references
  6. Generate pruning plan with specific actions (remove, condense, archive, preserve)
  7. Apply pruning according to mode (conservative/aggressive/dry-run)
  8. Update frontmatter with last-pruned date and reduction percentage
  9. Validate result to ensure document still valid after pruning
  10. Generate pruning report with summary and actions taken

For detailed implementation steps, see supporting documentation below.

Supporting Documentation

When you need detailed information, load the appropriate supporting file:

For Detailed Workflow

See instructions.md for:

  • Complete step-by-step pruning workflow
  • Historical content detection algorithms
  • Staleness scoring formulas and categories
  • Pruning plan generation
  • Pruning execution (conservative, aggressive, dry-run)
  • Frontmatter updates
  • Validation procedures
  • Report generation
  • Edge case handling (cross-references, partial historical, conflicting dates)

Load when: Performing pruning or need implementation details

For Pruning Strategies

See pruning-strategies.md for:

  • Remove entirely strategy (when and how)
  • Condense to summary strategy
  • Archive strategy
  • Preserve with marker strategy
  • Strategy selection criteria
  • Safety checks and rollback

Load when: Need strategy details or deciding which approach to use

For Content Patterns

See content-patterns.md for:

  • Historical content markers (keywords, dates, strikethrough)
  • Update history patterns
  • Deprecated section patterns
  • Redundant content patterns
  • Pattern matching algorithms
  • False positive handling

Load when: Detecting historical content or understanding pattern matching

For Usage Examples

See examples.md for:

  • Basic conservative pruning
  • Aggressive pruning
  • Dry-run preview
  • Pruning specific document
  • Example pruning reports (before/after)
  • Error scenarios (over-pruning, broken references)

Load when: User needs examples or clarification

Error Handling

No Historical Content Found

INFO: No historical content detected

Document: {doc}
All content appears current.

No pruning needed.

Over-Pruning Warning

WARNING: Pruning would remove >50% of content

Document: {doc}
Removal: {percentage}%

Suggestion: Review document - consider if it should be archived entirely
or switch to conservative mode.

Broken References After Pruning

ERROR: Pruning would break cross-references

Section: {section}
Referenced by: {other-docs}

Action: Preserved section to maintain references.
Update referencing docs before pruning.

Integration

Works well with:

  • /design-sync - Sync docs before pruning to ensure currency
  • /design-review - Identify documents needing pruning
  • /design-validate - Validate after pruning
  • /design-archive - Archive documents that are mostly historical

Success Criteria

A successful pruning:

  • ✅ Historical content identified accurately
  • ✅ Staleness scores calculated
  • ✅ Appropriate pruning strategy applied
  • ✅ Cross-references preserved
  • ✅ Document still validates
  • ✅ Frontmatter updated with pruning metadata
  • ✅ Clear pruning report generated
  • ✅ Content reduction documented

スコア

総合スコア

60/100

リポジトリの品質指標に基づく評価

SKILL.md

SKILL.mdファイルが含まれている

+20
LICENSE

ライセンスが設定されている

+10
説明文

100文字以上の説明がある

0/10
人気

GitHub Stars 100以上

0/15
最近の活動

3ヶ月以内に更新がある

0/10
フォーク

10回以上フォークされている

0/5
Issue管理

オープンIssueが50未満

+5
言語

プログラミング言語が設定されている

+5
タグ

1つ以上のタグが設定されている

0/5

レビュー

💬

レビュー機能は近日公開予定です