project-context-manager

Executes bash commands, modifies files, and runs startup scripts based on prompt patterns.

Defined in hooks/hooks.json

{
  "Stop": [
    {
      "hooks": [
        {
          "type": "command",
          "command": "bash $CLAUDE_PLUGIN_ROOT/hooks/scripts/save-state.sh",
          "timeout": 5
        }
      ],
      "matcher": "*"
    }
  ],
  "PreToolUse": [
    {
      "hooks": [
        {
          "type": "prompt",
          "prompt": "You are a credential leak prevention system. Analyze the tool use for potential credential exposure.\n\nCheck for these patterns in $TOOL_INPUT:\n1. **API keys/tokens**: Patterns like 'sk-...', 'ghp_...', long alphanumeric strings (>20 chars)\n2. **Private keys**: BEGIN RSA/OPENSSH/PGP PRIVATE KEY blocks\n3. **Passwords**: password=, PASSWORD=, pwd= followed by values\n4. **Credentials in files**: Writing to .env, credentials.json, secrets.yaml\n5. **Hardcoded secrets**: Assignment statements with suspicious values\n6. **Database credentials**: Connection strings with passwords\n\n**Analysis:**\n- If $TOOL_NAME is 'Write' or 'Edit': Check file_path and new_string/content for patterns\n- If $TOOL_NAME is 'Bash': Check command for exposed credentials\n- Ignore test files, example files, and fixture directories\n- Ignore commented-out code\n\n**Response format:**\n```json\n{\n  \"hookSpecificOutput\": {\n    \"permissionDecision\": \"allow|deny|ask\"\n  },\n  \"systemMessage\": \"Explanation for Claude\"\n}\n```\n\n**Return:**\n- \"allow\" if no credential patterns detected\n- \"deny\" with explanation if clear credential exposure found\n- \"ask\" if suspicious but not certain (let Claude decide)\n\n**Examples of DENY:**\n- File write to production.env with API_KEY=sk-...\n- Bash command with password in cleartext: curl -u admin:password123\n- Edit adding hardcoded token: TOKEN = \"ghp_abc123def456\"\n\n**Examples of ALLOW:**\n- Comments about credentials\n- Variable names without values: API_KEY=$API_KEY\n- Credential retrieval: /get-credential github-api\n- Test fixtures with fake credentials\n\nAnalyze now:",
          "timeout": 15
        }
      ],
      "matcher": "Bash|Edit|Write"
    }
  ],
  "SessionStart": [
    {
      "hooks": [
        {
          "type": "command",
          "command": "bash $CLAUDE_PLUGIN_ROOT/hooks/scripts/session-start.sh",
          "timeout": 10
        }
      ],
      "matcher": "*"
    }
  ]
}

{
  "riskFlags": {
    "touchesBash": true,
    "matchAllTools": false,
    "touchesFileWrites": true
  },
  "typeStats": {
    "prompt": 1,
    "command": 2
  },
  "eventStats": {
    "Stop": 1,
    "PreToolUse": 1,
    "SessionStart": 1
  },
  "originCounts": {
    "absolutePaths": 0,
    "pluginScripts": 0,
    "projectScripts": 0
  },
  "timeoutStats": {
    "commandsWithoutTimeout": 0
  }
}