OpenLang

Compact AI-to-AI communication protocol. Use when spawning sub-agents, sending inter-agent messages via sessions_send/sessions_spawn, or when instructed to s...

This skill is a message-format specification — it doesn't install code or ask for credentials, so it's internally coherent. However, OpenLang is expressly designed to encode commands targeting files, shells, networks, databases, and environment data; when used it can concisely express actions that, if executed by downstream agents, could access or exfiltrate sensitive data. Before enabling or using this skill: (1) ensure any sub-agents that receive OpenLang messages have limited, audited capabilities; (2) avoid using OpenLang on human-facing channels (the SKILL.md warns this); (3) consider logging, validating, or whitelisting OpenLang messages before execution to prevent hidden or compressed commands from performing unintended operations.

Type: OpenClaw Skill Name: openlang Version: 0.2.0 The skill defines 'OpenLang,' a compact domain-specific language (DSL) for agent-to-agent communication that includes sigils and vocabulary for high-risk operations such as shell execution (@sh), filesystem manipulation (@fs), and network access (@net). While the stated goal in SKILL.md is token reduction, the protocol functions as an obfuscation layer that could hide complex or malicious commands from human oversight. The instructions command the agent to use this protocol by default for all inter-agent communication, which significantly increases the attack surface for multi-agent prompt injection and makes the agent's internal logic harder to audit.