菜单
Install, configure, and manage the Claude Code Damage Control security hooks system. Use when user mentions damage control, security hooks, protected paths, blocked commands, install security, or modify protection settings.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
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/disler/claude-code-damage-control --skill damage-control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。
正在显示 SKILL.md
| name | Damage Control |
| description | Install, configure, and manage the Claude Code Damage Control security hooks system. Use when user mentions damage control, security hooks, protected paths, blocked commands, install security, or modify protection settings. |
Defense-in-depth protection system for Claude Code. Blocks dangerous commands and protects sensitive files via PreToolUse hooks.
This skill helps users deploy and manage the Damage Control security system, which provides:
ask: true)zeroAccessPaths - No access at all (secrets/credentials)readOnlyPaths - Read allowed, modifications blockednoDeletePaths - All operations except delete.claude/skills/damage-control/
├── SKILL.md # This file
├── patterns.yaml # Shared security patterns (single source of truth)
├── cookbook/
│ ├── install_damage_control_ag_workflow.md
│ ├── modify_damage_control_ag_workflow.md
│ ├── manual_control_damage_control_ag_workflow.md
│ ├── list_damage_controls.md
│ ├── test_damage_control.md
│ └── build_for_windows.md
├── hooks/
│ ├── damage-control-python/ # Python/UV implementation
│ │ ├── bash-tool-damage-control.py
│ │ ├── edit-tool-damage-control.py
│ │ ├── write-tool-damage-control.py
│ │ ├── python-settings.json
│ │ └── test-damage-control.py
│ └── damage-control-typescript/ # Bun/TypeScript implementation
│ ├── bash-tool-damage-control.ts
│ ├── edit-tool-damage-control.ts
│ ├── write-tool-damage-control.ts
│ ├── typescript-settings.json
│ └── test-damage-control.ts
└── test-prompts/ # Test prompts for validation
├── sentient_v1.md
├── sentient_v2.md
├── sentient_v3.md
└── sentient_v4.md
The install workflow copies hooks and creates settings based on the chosen level:
~/.claude/
├── settings.json # Hook configuration
└── hooks/
└── damage-control/
├── patterns.yaml
├── bash-tool-damage-control.py (or .ts)
├── edit-tool-damage-control.py
└── write-tool-damage-control.py
<agents current working directory>/
└── .claude/
├── settings.json # Hook configuration (shared)
└── hooks/
└── damage-control/
├── patterns.yaml
├── bash-tool-damage-control.py (or .ts)
├── edit-tool-damage-control.py
└── write-tool-damage-control.py
<agents current working directory>/
└── .claude/
├── settings.local.json # Personal overrides (gitignored)
└── hooks/
└── damage-control/
├── patterns.yaml
├── bash-tool-damage-control.py (or .ts)
├── edit-tool-damage-control.py
└── write-tool-damage-control.py
This section defines the decision tree for handling user requests. Based on what the user says, read and execute the appropriate workflow prompt.
Trigger phrases: "install damage control", "setup security hooks", "deploy damage control", "add protection"
Workflow: Read and execute cookbook/install_damage_control_ag_workflow.md
Trigger phrases: "help me modify damage control", "update protection", "change blocked paths", "add restricted directory"
Workflow: Read and execute cookbook/modify_damage_control_ag_workflow.md
Trigger phrases: "how do I manually update", "explain damage control config", "show me the settings"
Workflow: Read and execute cookbook/manual_control_damage_control_ag_workflow.md
Trigger phrases: - "test damage control", - "run damage control tests", - "verify hooks are working" - "damage control test this command " - "damage control test this read to this path " - "damage control test this write to this path " - "damage control test this delete to this path " - "damage control test this run this command "
Workflow: Read and execute cookbook/test_damage_control.md
What it does:
Trigger phrases: "build for windows", "add windows patterns", "convert to windows", "windows damage control"
Workflow: Read and execute cookbook/build_for_windows.md
What it does:
Trigger phrases: "update global read only paths to include X", "add /secret to zero access paths", "block command Y"
Action: Execute immediately without prompts - the user knows the system.
Examples:
| Level | Path | Scope |
|---|---|---|
| Global | ~/.claude/settings.json | All projects |
| Project | .claude/settings.json | Current project (shared) |
| Project Personal | .claude/settings.local.json | Current project (personal) |
| Type | Read | Write | Edit | Delete | Use Case |
|---|---|---|---|---|---|
zeroAccessPaths | No | No | No | No | Secrets, credentials |
readOnlyPaths | Yes | No | No | No | System configs, history |
noDeletePaths | Yes | Yes | Yes | No | Important project files |
| Implementation | Runtime | Install Command |
|---|---|---|
| Python | UV (Astral) | curl -LsSf https://astral.sh/uv/install.sh | sh |
| TypeScript | Bun | curl -fsSL https://bun.sh/install | bash && bun add yaml |
| Code | Meaning |
|---|---|
| 0 | Allow operation |
| 0 | Ask (JSON output triggers dialog) |
| 2 | Block operation |
Use the test prompts in test-prompts/ to validate the hooks:
sentient_v1.md - Tests rm -rf blocking (bashToolPatterns)sentient_v2.md - Tests find -delete blocking (noDeletePaths)sentient_v3.md - Tests ask patterns (SQL DELETE with ID)sentient_v4.md - Tests simple command blockingRun a test:
/project:test-prompts/sentient_v1
基于 SOC 职业分类