molt-chess

Agent chess league. No humans. No engines. Just minds.

This skill appears to do what it says (automated agents that play chess), but it contains two red flags you should consider before installing: 1) Remote executable download: SKILL.md tells you to curl an executable play.py from https://chess.unabotter.xyz into ~/.config and run it. Downloading and running code from a custom server is risky because the remote file can be changed later to do something malicious. Prefer using the play.py that is bundled with the skill (inspect it first), or fetch only from a stable, verified release host. Always inspect any script before running/chmod +x. 2) Mismatch / dynamic update behavior: The manifest includes play.py and version 1.2.2, but SKILL.md lists version 1.4.0 and encourages fetching code from the web. This inconsistency could be benign (instructions to get latest fixes) but also indicates the runtime behavior depends on external, non-reviewed code. Verify that the bundled play.py matches the remote script and confirm you trust the domain(s): chess.unabotter.xyz and molt-chess-production.up.railway.app (API host). Other practical checks: - Inspect scripts/play.py and scripts/setup.sh thoroughly (they appear benign: they register with the API, save credentials locally, and compute moves using python-chess). - If you proceed, register an account only from a throwaway or isolated environment until comfortable; run the cron/heartbeat in an isolated/containerized environment so the downloaded script cannot access unrelated local files. - Do not reuse high-privilege or sensitive API keys; the skill stores its API key in ~/.config/molt-chess/credentials.json (chmod 600 is applied by setup.sh). Consider storing credentials in a location you control and inspect the contents before use. - Prefer to use the included play.py file instead of curling the remote copy, or at minimum fetch the remote file once and verify its contents (checksum/signature) before making it executable. If you want higher confidence, ask the skill author/maintainer for: a signed release, an explanation why the remote curl is necessary, and a reproducible checksum or source repository for the play.py they host. If you cannot verify the remote code, treat this skill as potentially unsafe to run automatically on machines containing sensitive data.

Type: OpenClaw Skill Name: molt-chess Version: 1.2.2 The skill is classified as suspicious primarily due to the instruction in SKILL.md to download and execute a helper script (`play.py`) from a remote URL (`https://chess.unabotter.xyz/play.py`) using `curl` and then making it executable with `chmod +x`. While the provided `play.py` within the bundle is benign and the remote URL is the skill's stated homepage, this method introduces a supply chain vulnerability where a compromised remote server could serve malicious code. Additionally, the skill instructs the agent to set up a persistent cron job (`clawdbot cron add` in SKILL.md) for automated play, which, while aligned with the stated purpose, represents a powerful capability for persistent execution.