Openclaw Agent Protocol
Security checks across malware telemetry and agentic risk
Overview
The code, instructions, and requested resources line up with an OpenClaw gateway plugin that enforces routing; nothing in the package indicates covert exfiltration or unrelated privilege requests, though there are small coherence/details to review before installing.
This plugin appears to do what it claims: enforce routing at the OpenClaw gateway and log blocked calls locally. Before installing: (1) review and edit ~/.openclaw/hard_stop_policy.json immediately after setup so you control which tools are routed/blocked; (2) be aware the plugin will read ~/.openclaw/openclaw.json and will write logs and a local SQLite DB (~/.openclaw/compliance.db) containing truncated tool details (commands, paths, code up to ~200 chars) — these are stored locally only; (3) note small naming inconsistencies in the project (the installer/wizard uses "agent-routing-enforcer" while the plugin id in code is "openclaw-agent-protocol") — confirm your OpenClaw gateway will load the expected manifest and extension path; (4) run the setup in a test environment first and inspect the created files and plugin manifest before enabling in production; (5) run standard supply-chain checks (npm audit, check package origin, verify git repo) and ensure you have a build toolchain available for native deps (better-sqlite3). If you want, provide your OpenClaw gateway plugin-loading behavior (how it maps manifest IDs to installed directories) and I can point out exactly which filenames/IDs to inspect or adjust before enabling.
SkillSpector
SkillSpector findings are pending for this release.
VirusTotal
No VirusTotal findings