Eagle Tag Governance — Git skill for Claude Code

Eagle Tag Governance

English | [简体中文](readme/README.zh-CN.md)

An [Agent Skills](https://agentskills.io)-standard skill that governs the **tag vocabulary** of an [Eagle](https://eagle.cool/) library via the `eagle-mcp` MCP server — merging, renaming, normalizing, deduplicating, and retiring existing tags so the taxonomy stays clean and consistent. It runs in any Agent Skills-compatible AI agent (Claude Code, Codex CLI, Gemini CLI, GitHub Copilot, Cursor, WorkBuddy, …).

What it does

Instead of looking at individual assets, this skill works on the *tag list itself*:

1. Audit — lists every tag with its usage count (tag_get / tag_count).
2. Detect — spots synonyms, spelling/case/space variants, low-frequency noise (count = 0/1), and tags that drift away from the controlled three-dimension taxonomy.
3. Plan — produces an editable merge_plan.json (merges / renames / retire).
4. Preview & authorize — surfaces every operation with its blast radius; you review and confirm (and nothing is written without your okay).
5. Execute — applies tag_merge / tag_update safely.
6. Verify — reads the tags back to confirm the source is gone and counts are correct.

Highlights

1. Safe by design — every write is dry-run first, then authorized. Tag merges/renames are global and irreversible, so the skill never skips the preview or the confirmation gate.
2. Undo mapping — exports tag_undo_mapping.json (source→target / old→new) before any write, so a bad merge can be manually reversed.
3. Normative target — judges "drift" against the same controlled three-dimension vocabulary (vocabulary*.md) used by eagle-untagged-organizer, keeping both skills aligned.
4. Honest about limits — Eagle's MCP has no tag_delete; retiring a tag can only merge it or strip it from items (full deletion is a manual UI step), and the plan says so plainly.

Install

Install the skill folder into your AI agent's skills directory:

| Agent |