The Agent Skills Directory

[COMMAND_EXECUTION] (MEDIUM): The skill uses a Node.js wrapper (setup-cloudbase-openclaw.js) to execute internal scripts via spawnSync. It also references a shell script scripts/detect-setup.sh in the README. These scripts are intended to modify system-level configurations (e.g., ~/.openclaw/openclaw.json, mcporter.json).
[EXTERNAL_DOWNLOADS] (MEDIUM): The skill encourages the installation of packages and plugins from the @cloudbase npm scope and the tencentcloudbase/skills GitHub repository. While likely legitimate for Tencent Cloud users, this organization is not on the 'Trusted GitHub Organizations' whitelist, requiring manual verification of the downloaded content.
[PROMPT_INJECTION] (LOW): The 'Skill Enhancer' plugin (plugins/skill-enhancer/index.ts) programmatically injects instructions into the AI agent's prompt during the before_agent_start event. It uses authoritative language ('You MUST', 'NEVER skip') to override the model's default reasoning and force specific skill usage patterns.
[DATA_EXPOSURE] (SAFE): The skill provides templates for storing cloud provider credentials (TENCENTCLOUD_SECRETID, TENCENTCLOUD_SECRETKEY) in local JSON configuration files. While this is standard for MCP setups, users should be aware that these secrets are stored in plaintext on the filesystem.
[UNVERIFIABLE_CODE] (MEDIUM): Key logic files referenced in package.json and README.md, specifically scripts/setup.mjs, scripts/detect-setup.sh, and scripts/test-env.mjs, are missing from the provided source code. This prevents a full security audit of the file system modifications and network operations performed during the setup process.

setup-cloudbase-openclaw