Universal Skills Manager

The master coordinator for AI skills. Discovers skills from multiple sources (SkillsMP.com, SkillHub, and ClawHub), manages installation, and synchronization...

This skill appears to implement a legitimate universal skills manager, but there are several red flags you should consider before installing or giving credentials: - Inconsistencies: The registry metadata does not declare the SKILLSMP_API_KEY or required binaries, yet SKILL.md and config.json do. Ask the publisher to reconcile metadata and frontmatter so you know exactly what will be requested. - Review code before running: The package includes installer and downloader scripts that will fetch and write arbitrary files into many ~/.tool/ and project directories. Inspect scripts/install_skill.py and scripts/scan_skill.py yourself or run them in a sandboxed VM/container first. - Never pipe unknown remote scripts to sh: The SKILL.md recommends curl | sh as a convenience — avoid this. If you must run an installer from the project, fetch it, inspect it locally, then run it. - Limit credentials: Only provide SKILLSMP_API_KEY or GitHub tokens if you trust the publisher and understand scope; consider creating scoped or temporary tokens with minimal privileges. - Test safely: If you want to try the skill, do so on a disposable account or isolated environment (container or VM) and verify exactly which paths it modifies. If the publisher can (1) update registry metadata to declare required env vars and binaries, (2) remove or document the curl|sh shortcut, and (3) publish a verifiable homepage/release with checksums, the risk profile would be clearer and easier to evaluate.

Type: OpenClaw Skill Name: universal-skills-manager Version: 1.7.0 The 'universal-skills-manager' bundle is a comprehensive tool for discovering, installing, and synchronizing AI agent skills across multiple platforms. It possesses high-risk capabilities, specifically the ability to download and execute remote scripts from GitHub and ClawHub, and perform broad file system operations across various AI tool directories (e.g., ~/.claude, ~/.gemini). While it includes a dedicated security scanner (scripts/scan_skill.py) and validation logic to mitigate risks, the inherent potential for remote code execution (RCE) and the inclusion of 'curl | sh' installation instructions in SKILL.md for external tools warrant a suspicious classification despite the behavior being aligned with the stated purpose.