AgentSkillsCN

Content Deduplicator

内容去重器

SKILL.md

Content Deduplicator Skill

Purpose

This skill provides a systematic approach to reducing documentation debt without losing technical research. It focuses on Merging and Preserving micro-scenarios into authoritative technical whitepapers.

Philosophy: Merge & Preserve

  1. •Never delete unique insights: If a small file contains a browser version or a specific reproduction step not in the parent, MOVE it, don't drop it.
  2. •Authoritative Parents: All micro-scenarios must point to a "Mega-Whitepaper" (Interaction, UI, Performance, etc.).
  3. •Surgical Consolidation:
    • •Identify semantic overlap.
    • •Append unique sections of the child to the parent.
    • •Update all case files (scenarioId) before considering child deletion.

Workflow

1. structural & logical Audit

Before merging content, perform a deep audit to detect non-semantic duplication:

  • •Agent-Led Manual Comparison: Select a candidate file and surgically compare its technical sections (Phenomenon, Cause, Solution) against other files in the same category. Focus on technical nuance rather than just text overlap.
  • •Cross-File Shadowing: Search for exact technical snippets (e.g., same code blocks) across different file categories.

2. The Merge Procedure

  • •Context: Copy the "Phenomenon" or "Cause" from the micro-document.
  • •Integration: Add a new sub-header in the Mega-Whitepaper or integrate into an existing one.
  • •References: Move the "Related Cases" links to the Mega-Whitepaper's reference section.
  • •Normalization: When merging bilingual content, ensure both EN and KO versions are updated simultaneously to maintain parity.

3. File Decommissioning

ONLY after verifying that:

  1. •All content exists in the parent.
  2. •All cases point to the parent.
  3. •The parent build is valid (zero "Duplicate ID" warnings).

mega-whitepaper registry

  • •scenario-ime-interaction-patterns: Functional keys, beforeinput sequences, composition start/end, auto-input behaviors.
  • •scenario-ime-ui-experience: Viewport, keyboard, candidate windows, scrolling, anchoring.
  • •scenario-ime-language-specifics: CJK, RTL, Thai, Indic unique behaviors.
  • •scenario-performance-foundations: RuleSet invalidation, memory leaks, large doc lag, CSS complexity.
  • •scenario-accessibility-foundations: ARIA-Tree, screen readers, roles, spellcheck interference.
  • •scenario-content-normalization: Paste, whitespace, DOM hygiene,   management.