Omniclaw Cli Skill

Use this skill whenever an agent needs to pay for an x402 URL, transfer USDC to an address, inspect OmniClaw balances or ledger entries, or explicitly expose...

This skill appears to do what it says: operate the omniclaw-cli for payments, balance inspection, and (with explicit owner approval) exposing paid endpoints. Before installing, ensure: (1) the omniclaw-cli binary is obtained from a trusted source and up-to-date; (2) only provide OMNICLAW_TOKEN and OMNICLAW_SERVER_URL that you trust (do not supply OMNICLAW_OWNER_TOKEN unless you intentionally allow owner approval actions); (3) block or tightly control use of omniclaw-cli serve/--exec unless an owner explicitly approves the exact command and you run it in an isolated container or private network; (4) avoid running the included docs generator scripts in an environment where live tokens are present unless you trust the CLI help output — the generator forwards your environment to child processes; and (5) add monitoring/limits for payments so accidental or unauthorized spends are detectable. If you need higher assurance, request the omniclaw-cli source or a reproducible package verification from the publisher before granting tokens.

Type: OpenClaw Skill Name: omniclaw Version: 0.0.8 The skill bundle is classified as suspicious due to the high-risk 'serve --exec' functionality in the omniclaw-cli, which allows for host command execution triggered by network requests. While SKILL.md includes defensive instructions and requires owner approval, the inherent risk of this feature is significant. Additionally, scripts/generate_cli_reference.py contains a suspicious path calculation (parents[3]) that attempts to write documentation to a directory outside the bundle's root, and it executes the omniclaw-cli binary from the environment, which could be exploited if the binary is untrusted.