sync-trending

Monitior technology trends (GitHub, etc.), contextualize them against the user's project, and autonomously verify them through installation and testing. Use...

This skill does what it says — it will look at your local project and, if you approve, clone and run third‑party repositories to verify them. That is powerful but risky: cloning and installing dependencies can execute arbitrary code (via postinstall scripts or build steps) and might access or leak data. Before using this skill, consider the following: 1) Only allow deep dives when you trust the repository and explicitly grant permission each time. 2) Prefer the skill be changed to use strong isolation (run clones inside a disposable VM or container, drop privileges, and avoid cloning into your home directory). 3) Require and enforce dependency isolation (Python virtualenv/venv, use npm ci with --ignore-scripts or install in a container) to reduce postinstall risk. 4) Never let it access secrets or environment variables; verify the agent's memory ('save_memory') does not contain sensitive data. 5) If you need stronger guarantees, run manual verifications yourself in an isolated environment or ask the skill author to add explicit sandboxing instructions. Providing the skill with explicit sandbox/container steps or a policy that forbids global installs would reduce my concern.

Type: OpenClaw Skill Name: sync-trending Version: 0.1.0 This skill is classified as suspicious due to its inherent design to execute arbitrary, untrusted code from external sources. The `SKILL.md` explicitly instructs the agent to use `run_shell_command` to `git clone` repositories and then run build/install commands (`npm install`, `pip install`, `npm start`) based on instructions found in the untrusted repository's `README.md`. While the skill includes safety guidelines like requiring explicit user permission, cloning into a temporary directory, and instructing the agent not to expose secrets, the act of executing code from untrusted trending repositories (a known supply chain attack vector) poses a significant risk of remote code execution if a trending project is malicious or compromised. This represents a critical vulnerability rather than intentional malice by the skill itself.