Capability inventory & security audit tool for Claude Code
Scan all installed plugins, user commands, agents, and skills — generate a structured capability inventory, risk report, and change detection.
# 1. Add the marketplace
/plugins marketplace add nestedcat/nestedcat-claude-marketplace
# 2. Install the plugin
/plugins install plugin-audit@nestedcat-claude-marketplace| Command | Description |
|---|---|
/audit |
Full capability scan with cache health report |
/audit-help <name> |
Detailed info for a specific capability |
/audit-risk |
Risk assessment report for all capabilities |
/audit-diff |
Compare current state against a saved snapshot |
- Plugins:
~/.claude/plugins/installed_plugins.json→ each plugin's commands, skills, agents, hooks - User-level:
~/.claude/commands/,agents/,skills/ - Project-level:
.claude/commands/,agents/,skills/
Running /audit produces a structured capability inventory like this:
| Name | Type | Source | Provider | Invocation |
|---|---|---|---|---|
| frontend-design | skill | plugin:frontend-design | claude-plugins-official | Auto-triggered by model |
| setup_semgrep_plugin | command | plugin:semgrep-plugin | claude-plugins-official | /setup_semgrep_plugin |
| code-review | command | plugin:code-review | claude-plugins-official | /code-review |
| code-simplifier | agent | plugin:code-simplifier | claude-plugins-official | Use code-simplifier agent |
| revise-claude-md | command | plugin:claude-md-management | claude-plugins-official | /revise-claude-md |
| audit | command | plugin:plugin-audit | nestedcat-claude-marketplace | /audit |
| PostToolUse:Write|Edit | hook | plugin:semgrep-plugin | claude-plugins-official | Auto-trigger |
| planner | agent | user | — | Use planner agent |
| tdd-guide | agent | user | — | Use tdd-guide agent |
Statistics: 18 capabilities — 6 commands, 3 skills, 3 agents, 6 hooks · from 6 plugins + user-level
Notable findings:
claude-md-improverhas broad permissions includingBashsemgrep-pluginregisters 4 hooks that auto-trigger on tool use and session start
The report also includes a Cache Health section showing orphaned plugin versions, version conflicts, and blocked plugins.
> /audit-help code-review
# Shows detailed info: description, permissions, invocation, source path
> /audit-risk
# Shows risk scores: CRITICAL / HIGH / MEDIUM / LOW
> /audit-diff
# Shows added / removed / changed capabilities since last snapshot
| Permission | Score |
|---|---|
| Bash (unrestricted) | 15 |
| Bash (pattern) | 8 |
| Write | 6 |
| Edit / NotebookEdit | 5 |
| WebFetch | 4 |
| Task / WebSearch | 3 |
| Read | 2 |
| Glob / Grep | 1 |
| Hook (auto-trigger) | +3 |
| Hook (wildcard) | +2 |
Levels: CRITICAL (≥15) · HIGH (≥10) · MEDIUM (≥5) · LOW (<5)
- Claude Code with Plugin support
- Python 3.11+ (stdlib only, zero dependencies)
# Run tests
cd scripts && python3 -m unittest test_scan -v
# Run scan directly
python3 scripts/scan.py --all
python3 scripts/scan.py --risk
python3 scripts/scan.py --detail <name>
# Load plugin locally
claude --plugin-dir /path/to/plugin-auditThis tool provides automated capability scanning and risk scoring as a convenience. It is not a substitute for professional security auditing. Risk scores are heuristic-based and may not capture all threats. Use at your own discretion.
Claude Code 能力索引与安全审计工具
扫描所有已安装的插件、用户命令、Agent 和 Skill，生成结构化能力清单、风险报告和变更检测。
# 1. 添加插件市场
/plugins marketplace add nestedcat/nestedcat-claude-marketplace
# 2. 安装插件
/plugins install plugin-audit@nestedcat-claude-marketplace| 命令 | 说明 |
|---|---|
/audit |
全面扫描，输出能力清单和缓存健康报告 |
/audit-help <name> |
查询单个能力的详细信息 |
/audit-risk |
所有能力的风险评估报告 |
/audit-diff |
与历史快照对比，检测变更 |
- 插件:
~/.claude/plugins/installed_plugins.json→ 每个插件的命令、技能、Agent、Hook - 用户级:
~/.claude/commands/、agents/、skills/ - 项目级:
.claude/commands/、agents/、skills/
运行 /audit 后，Claude 会将扫描结果整理为如下能力清单表格：
| 名称 | 类型 | 来源 | 提供者 | 调用方式 |
|---|---|---|---|---|
| frontend-design | skill | plugin:frontend-design | claude-plugins-official | 模型自动触发 |
| setup_semgrep_plugin | command | plugin:semgrep-plugin | claude-plugins-official | /setup_semgrep_plugin |
| code-review | command | plugin:code-review | claude-plugins-official | /code-review |
| code-simplifier | agent | plugin:code-simplifier | claude-plugins-official | Use code-simplifier agent |
| revise-claude-md | command | plugin:claude-md-management | claude-plugins-official | /revise-claude-md |
| audit | command | plugin:plugin-audit | nestedcat-claude-marketplace | /audit |
| PostToolUse:Write|Edit | hook | plugin:semgrep-plugin | claude-plugins-official | 自动触发 |
| planner | agent | user | — | Use planner agent |
| tdd-guide | agent | user | — | Use tdd-guide agent |
统计: 18 个能力 — 6 命令、3 技能、3 Agent、6 Hook · 来自 6 个插件 + 用户级
重要发现：
claude-md-improver拥有包括Bash在内的广泛权限semgrep-plugin注册了 4 个 Hook，在工具调用和会话启动时自动触发
报告还包含 缓存健康 部分，显示孤立的插件版本、版本冲突和被封锁的插件。
> /audit-help code-review
# 显示详细信息：描述、权限、调用方式、源文件路径
> /audit-risk
# 显示风险评分：CRITICAL / HIGH / MEDIUM / LOW
> /audit-diff
# 显示新增 / 删除 / 变更的能力（与上次快照对比）
| 权限 | 分值 |
|---|---|
| Bash（无限制） | 15 |
| Bash（受限模式） | 8 |
| Write | 6 |
| Edit / NotebookEdit | 5 |
| WebFetch | 4 |
| Task / WebSearch | 3 |
| Read | 2 |
| Glob / Grep | 1 |
| Hook（自动触发） | +3 |
| Hook（通配匹配） | +2 |
风险等级: CRITICAL (≥15) · HIGH (≥10) · MEDIUM (≥5) · LOW (<5)
- 支持插件的 Claude Code
- Python 3.11+（仅标准库，零外部依赖）
# 运行测试
cd scripts && python3 -m unittest test_scan -v
# 直接运行扫描
python3 scripts/scan.py --all
python3 scripts/scan.py --risk
python3 scripts/scan.py --detail <name>
# 本地加载插件测试
claude --plugin-dir /path/to/plugin-audit本工具提供自动化的能力扫描和风险评分，仅作为辅助参考。它不能替代专业的安全审计。风险评分基于启发式规则，可能无法覆盖所有威胁。请自行判断后使用。