forge-security
为 Claude Code 强制执行安全防护。阻止访问机密、凭据和敏感文件。要求确认网络请求和基础设施变更。在访问文件、发出网络请求或运行基础设施命令时使用。
📖 中文导读:为 Claude Code 强制执行安全防护。阻止访问机密、凭据和敏感文件。要求确认网络请求和基础设施变更。在访问文件、发出网络请求或运行基础设施命令时使用。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。
name: forge-security description: Enforces security guardrails for Claude Code. Blocks access to secrets, credentials, and sensitive files. Requires confirmation for network requests and infrastructure changes. Use when accessing files, making network requests, or running infrastructure commands.
Security Guardrails
Blocked Operations (NEVER do these)
The following are blocked without exception:
- Reading or writing
.envfiles - Accessing files matching:
*secret*,*credential*,*password*,.aws/*,.ssh/*,*token* - Running destructive IaC commands without explicit user request
See reference/blocked-patterns.md for complete list.
Ask First (require user confirmation)
These operations require explicit user approval:
- Network requests (WebFetch)
- Git push, merge, rebase operations
- Infrastructure commands: terraform plan/apply, kubectl apply, ansible-playbook
See reference/ask-patterns.md for complete list.
When Uncertain
If unsure whether an operation is safe:
- Stop and explain what you're about to do
- List what files/systems will be affected
- Wait for explicit user confirmation
Never assume permission. When in doubt, ask.