用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/tomevault-io/skills-registry --skill ai-crap-manager命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
| Use when this capability is needed.
> Use when this capability is needed.
Review architecture and API design for the vfs-s3 project. Use when the user mentions @architect, asks to review an issue's design, discuss module boundaries, API shape, or architectural decisions for vfs-s3. Also trigger when the user wants to create an ADR (Architecture Decision Record) or evaluate a technical approach for the project. Intended for dispatch from Codex automation or Claude routines; GitHub trigger phrase: @vfs-s3-bot please prepare design doc Use when this capability is needed.
基于 SOC 职业分类
正在显示 SKILL.md
| name | ai-crap-manager |
| description | | Use when this capability is needed. |
Centralized management for AI-generated configuration files across all your projects.
AI coding tools (Claude Code, Cursor, Windsurf, Copilot, etc.) generate config directories and files at project roots:
.claude/, CLAUDE.md, .claude/settings.json.cursor/, .cursorules.windsurf/, .windsurfrules.github/copilot-instructions.md.aider*, .cline/, .roo/, .codex/, .bolt/, .devin/, .replit/These pollute git history, create merge conflicts, and clutter repos.
~/.agent-crap/ is the single source of truth. Each project gets a subdirectory. Files are symlinked back to project roots.
~/.agent-crap/
projects/
my-app/ # mirrors ~/Code/my-app AI configs
.claude/
CLAUDE.md
.cursor/
another-repo/
.claude/
CLAUDE.md
templates/ # reusable config templates
snapshots/ # named point-in-time snapshots
hooks/ # auto-sync watcher scripts
backups/ # automatic backups before operations
logs/ # hook execution logs
registry.json # maps project names -> original paths
ACM="bash ~/.claude/skills/ai-crap-manager/scripts/setup.sh"
# Or after alias-install:
# acm <command>
$ACM init # Create ~/.agent-crap/
$ACM register ~/Code/my-app # Move configs to store, symlink back
$ACM alias-install # Add 'acm' alias to shell
$ACM init # Initialize ~/.agent-crap/ store
$ACM register [path] # Register project (default: cwd)
$ACM unregister [path] # Restore files and remove from registry
$ACM sync [path] # Verify/fix symlinks, absorb new AI files
$ACM sync-all # Sync ALL registered projects
$ACM status [path] # Show tracked files, symlink health, snapshots
$ACM list # List all registered projects with metadata
$ACM detect [path] # Scan which AI tools are in use
$ACM gitignore [path] # Update .gitignore only
$ACM batch-register ~/Code/* # Register multiple projects at once
$ACM scan ~/Code # Scan directory for repos with AI configs
$ACM adopt ~/Code # Scan + register all found projects
scan shows what exists without changing anything. adopt does scan + register in one step.
Templates let you save and reuse AI configs across projects.
$ACM template list # List available templates
$ACM template save <name> [path] # Save project configs as template
$ACM template apply <name> [path] # Apply template to a project
$ACM template delete <name> # Delete a template
$ACM template show <name> # Inspect template contents
Use cases:
# Save your best setup
$ACM template save swift-pro ~/Code/my-polished-app
# Start a new project with it
$ACM template apply swift-pro ~/Code/new-app
Snapshots capture the exact state of a project's AI configs at a point in time. Unlike backups (automatic, time-based), snapshots are named and intentional.
$ACM snapshot <name> [path] # Create named snapshot
$ACM snapshot-list [path] # List snapshots for project
$ACM snapshot-restore <name> [path] # Restore a snapshot
Use cases:
$ACM snapshot before-refactor
# ... experiment with configs ...
$ACM snapshot-restore before-refactor # oops, go back
$ACM diff [path] # Show changes between store and project
$ACM tree [path] # Tree view of managed files
$ACM stats # Global statistics across all projects
diff catches files that were modified directly in the project instead of through symlinks (e.g., after a git checkout that replaced a symlink with a real file).
stats shows: project count, agent popularity, disk usage breakdown.
Track ANY file or directory beyond the built-in AI patterns. Useful for project-specific configs, prompt libraries, AI training data, or anything else you want centralized.
$ACM track <file-or-dir> [path] # Move to store + symlink + gitignore
$ACM untrack <file-or-dir> [path] # Restore to project + remove from store
$ACM tracked [path] # List all tracked files (built-in + custom)
Features:
$ACM track "file1,file2,dir1"$ACM track "*.prompts"$ACM track configs/ai/prompts.yamlExamples:
# Track a custom prompt library
$ACM track prompts/
# Track project-specific AI training data
$ACM track .ai-data,.ai-prompts
# Track a Makefile.ai and custom rules
$ACM track "Makefile.ai,rules.yaml"
# Stop tracking .cursorules (restore to project)
$ACM untrack .cursorules
# See what's tracked (split by built-in vs custom)
$ACM tracked
Install hooks so syncing happens automatically.
$ACM hook-install [path] # Install git hooks + optional fswatch/LaunchAgent
$ACM hook-remove [path] # Remove all installed hooks
What gets installed:
| Hook | Trigger | Purpose |
|---|---|---|
| Git post-checkout | Branch switch | Recreate symlinks after checkout |
| Git post-merge | After merge/pull | Absorb new AI files |
| fswatch watcher | File changes in store | Live sync (optional, if fswatch installed) |
| LaunchAgent (macOS) | WatchPaths on store dir | Background sync daemon |
$ACM doctor # Health check: registry, symlinks, orphans
$ACM gc # Remove old backups (>30 days), orphaned dirs
$ACM export <path.tar.gz> # Export entire store as portable archive
$ACM import <archive.tar.gz> # Import store from archive
$ACM alias-install # Add 'acm' alias to .zshrc/.bashrc
--dry-run # Show what would happen without doing it
--force # Skip confirmation prompts
--verbose # Detailed output
--only <agent> # Only track specific agent (claude, cursor, etc.)
| Pattern | Tool |
|---|---|
.claude/ | Claude Code |
CLAUDE.md | Claude Code |
.cursor/ | Cursor |
.cursorules | Cursor |
.windsurf/ | Windsurf |
.windsurfrules | Windsurf |
.github/copilot-instructions.md | GitHub Copilot |
.vscode/ai-instructions.md | VS Code AI |
.aider* | Aider |
.cline/ | Cline |
.clinerules | Cline |
.roo/ | Roo |
.roomodes | Roo |
.codex/ | Codex |
.copilot/ | Copilot |
.bolt/ | Bolt |
.devin/ | Devin |
.replit/ | Replit |
.idx/ | IDX |
$ACM init
$ACM adopt ~/Code # Finds and registers all repos with AI configs
$ACM alias-install # 'acm' shortcut
cd ~/Code/new-project
acm template apply swift-pro # Apply your standard AI config
acm hook-install # Auto-sync on git operations
acm snapshot stable # Save current working state
# ... edit CLAUDE.md, add hooks, change agents ...
acm snapshot-restore stable # Nope, go back
# Old machine:
acm export ~/Desktop/ai-configs.tar.gz
# New machine:
acm import ~/Desktop/ai-configs.tar.gz
acm sync-all # Recreate all symlinks
acm doctor # Check health
acm gc # Clean old backups
acm stats # See usage overview
~/.agent-crap/registry.json:
{
"version": 2,
"projects": {
"my-app": {
"path": "/Users/bruno/Code/my-app",
"registered": "2026-03-12T10:00:00+00:00",
"last_sync": "2026-03-12T15:30:00+00:00",
"tracked_files": [".claude/", "CLAUDE.md", ".cursor/"],
"agents": ["claude-code", "cursor"],
"tags": []
}
}
}
unregister restores files to their original locationssnapshot-restore backs up current state before restoringimport backs up existing store before overwritingdoctor catches corruption, orphans, and broken linksgc only removes backups older than 30 daysSource: brunogama/ai-crap-manager — distributed by TomeVault.