Content
<div align="center">
# AIDA
### AI Tool Asset Governance Layer — Rules, Skills, Decision Memory, Plugins
[](./package.json)
[](./LICENSE)
[](https://nodejs.org)
[](#verification)
[](https://glama.ai/mcp/servers/LWTlong/aida)
</div>
---
## Three Main Capabilities
### 1. Asset Governance
Scan directories like `.claude/rules/`, `.cursor/rules/`, `.codex/`, `.aida/`, etc., to identify duplicate, conflicting, or obsolete rules and skills. All write operations go through the `aida_apply_governance` MCP tool, with journaling for each batch, and can be rolled back with `aida_undo`.
**Core skill:** `/aida-govern` (all-in-one)
### 2. Layered Decision Memory
Record "why this code is written this way" in MADR format in `.claude/rules/decisions/`. Claude Code automatically loads relevant files via `paths` metadata when opened, no manual recall needed.
**Core skills:** `/aida-remember`, `/aida-remember-branch`
### 3. AI Asset Overview
Local Dashboard (`aida dashboard`) provides a package.json-style asset list: rules/skills/decision/MCP config/plugins, all in one screen.
**Core skill:** `/aida-audit`
---
## Get Started in 30 Seconds (Claude Code Plugin)
```
/plugin marketplace add LWTlong/aida
/plugin install aida@aida
```
Restart Claude Code. 12 `aida-*` skills are immediately available.
**Upgrade:**
```
/plugin update aida
```
### Cursor / Codex / VS Code (optional, via MCP)
Add to project root `.mcp.json`:
```json
{ "mcpServers": { "aida": { "command": "npx", "args": ["-y", "aida", "mcp"] } } }
```
Or use a one-click script:
```bash
npx aida init
```
---
## Skill Overview
| Category | Skill | Purpose |
|---|---|---|
| **Governance** | `/aida-govern` | All-in-one: scan → analyze → confirm → apply (covers rule overload/task-based rules/document redundancy/cross-tool duplicates/conflicts/aggregated mirrors/2.x legacy) |
| | `/aida-resolve` | Resolve git merge conflicts for AI asset files |
| | `/aida-sync` | Cross-tool sync: Claude ↔ Cursor ↔ Codex (select source → target) |
| | `/aida-import` | Securely import external Claude Plugin from local directory path (audit risks first) |
| **Read-only** | `/aida-audit` | Scan and organize all AI assets in the project |
| **Memory** | `/aida-remember` | Record a decision as MADR, automatically loaded by Claude Code |
| | `/aida-remember-branch` | Batch extract decisions from current branch diff |
| | `/aida-recall` | Recall previously沉淀 modules memory |
| | `/aida-analyze` | Record "what changed, why" after branch/function development |
| | `/aida-undo` | Roll back last AIDA write operation |
| **Sharing** | `/aida-pkg` | Package assets as Claude Plugin |
| | `/aida-ui` | Open local Dashboard (check npm package → guide installation → provide startup command) |
| | `/aida-help` | Display skill reference card |
---
## MCP Tool List
| Tool | Function |
|------|------|
| `aida_bootstrap` | Initialize / status check / tool authorization list |
| `aida_scan_assets` | Scan all AI assets |
| `aida_list_assets` / `aida_get_asset` | Browse assets |
| `aida_apply_governance` | Execute governance actions (create/modify/delete/remove-lines), all journaled |
| `aida_undo` | Revert last governance operation |
| `aida_write_analysis` | Write analysis report |
| `aida_memory` / `aida_remember` / `aida_recall` | Read/write decision memory |
| `aida_build_plugin` | Package project assets as Claude Plugin |
| `aida_build_self_plugin` | Package AIDA itself as Plugin |
| `aida_parse_plugin` / `aida_audit_plugin_risk` | Parse/audit external Plugin |
---
## Project Structure
```text
.claude-plugin/
plugin.json # Claude Plugin list (this repo itself is a plugin)
marketplace.json # Marketplace list
src/
assets/skills/ # 12 aida-* skill sources (plugin skills field points to)
core/ # scan, governance, undo, memory, plugin builder
mcp/server.ts # MCP server
cli/ # aida CLI
dashboard/ # Dashboard frontend output
.claude/
rules/decisions/ # MADR decision memory (Claude Code auto-loads)
.aida/
cache/ # scan results + undo journal
```
---
## Verification
```bash
npm test
npm run build
```
