AgentSkillsCN

authorize

为预设工具使用钩子绕过机制，持续维护白名单管理。

SKILL.md
--- frontmatter
name: authorize
description: Manage persistent allowlist for pre_tool_use hook bypass
triggers:
  - /authorize
  - authorize path
  - allow file access

Authorize

Manage the persistent allowlist at ~/.claude/authorized-paths.txt that bypasses pre_tool_use.py security checks.

Commands

CommandAction
/authorize add <pattern>Add path pattern to allowlist
/authorize remove <pattern>Remove pattern from allowlist
/authorize listShow current allowlist
/authorize clearClear all authorized paths

Usage

bash
# Allow a specific file
/authorize add ~/.secrets/api-keys.json

# Allow a directory pattern
/authorize add ~/.config/myapp/*

# Allow by extension
/authorize add *.local

# Remove authorization
/authorize remove ~/.secrets/api-keys.json

# View all authorized paths
/authorize list

Implementation

When user invokes /authorize:

add <pattern>

bash
echo "<pattern>" >> ~/.claude/authorized-paths.txt
sort -u ~/.claude/authorized-paths.txt -o ~/.claude/authorized-paths.txt

Confirm: "✓ Added <pattern> to authorized paths"

remove <pattern>

bash
grep -v "^<pattern>$" ~/.claude/authorized-paths.txt > /tmp/auth.tmp
mv /tmp/auth.tmp ~/.claude/authorized-paths.txt

Confirm: "✓ Removed <pattern> from authorized paths"

list

bash
cat ~/.claude/authorized-paths.txt 2>/dev/null || echo "(empty)"

Display as formatted list.

clear

bash
rm -f ~/.claude/authorized-paths.txt

Confirm: "✓ Cleared all authorized paths"

Pattern Matching

The hook uses these matching rules:

  • •Exact match: /path/to/file.txt
  • •Glob patterns: *.local, ~/.config/*
  • •Directory prefix: /path/to/dir/ (trailing slash = all children)

Security Notes

  • •Authorizations persist across sessions
  • •Review periodically with /authorize list
  • •Use narrow patterns (specific files > broad globs)
  • •.env files have built-in allowlist in hook (.env.local, .env.sample, .env.example)