X-Claw

Operate the local X-Claw agent runtime for intents, approvals, execution, reporting, and wallet operations.

This skill appears to be a full integration for a local X-Claw agent, but it performs invasive host changes and relaxes wallet policy defaults by design. Before installing: 1) Only proceed if you trust the source (https://xclaw.trade) and can verify release signatures or provenance. 2) Inspect scripts (setup_agent_skill.py and openclaw_gateway_patch.py) yourself — they will copy files into ~/.openclaw, create launchers on PATH, and patch OpenClaw JS bundles. 3) Do not run the installer on a machine with production wallets or real funds; prefer a disposable VM or test environment. 4) If you must install, set XCLAW_OPENCLAW_AUTO_PATCH=0 to disable automatic gateway patching, back up ~/.openclaw, and review/modify the default policy file (~/.xclaw-agent/policy.json) to require explicit approvals and lower spend caps. 5) Confirm required env vars (XCLAW_AGENT_API_KEY, XCLAW_API_BASE_URL, XCLAW_DEFAULT_CHAIN) are set appropriately and avoid supplying real wallet passphrases during initial setup. If you can obtain an official signed release or documentation from the project author confirming the patcher/installer steps, that would reduce risk; absent that, treat this as high-risk and review manually.

Type: OpenClaw Skill Name: x-claw Version: 0.1.0 The skill bundle is classified as suspicious due to its highly invasive installation and runtime modification practices. The `scripts/setup_agent_skill.py` script performs extensive system changes, including creating executable launchers on PATH, modifying OpenClaw's configuration, and crucially, executing `scripts/openclaw_gateway_patch.py`. This patcher script directly modifies OpenClaw's core JavaScript bundles (`reply-*.js`) to inject custom logic for handling Telegram approvals, a high-privilege operation that alters the host environment's functionality. While the stated intent of these modifications is to enhance security by bypassing the LLM for critical approval decisions and improving UX, these methods introduce a significant attack surface and potential for critical vulnerabilities (e.g., RCE if the patched code or the `xclaw-agent` binary it calls were compromised), even if no explicit malicious payload is present in the current version.