openlang

What This Skill Does

OpenLang is a highly optimized, compact communication protocol designed specifically for agent-to-agent interactions within the OpenClaw ecosystem. By using sigils, abbreviated action tokens, and a structured shorthand, OpenLang reduces message token usage by 5-10x compared to standard natural language. This efficiency is critical when sub-agents are spawned, as it reduces the overhead in context windows and lowers latency in multi-agent orchestration. It functions as a specialized "machine language" for agents, enabling high-density information exchange while maintaining semantic clarity for any agent implementation that supports the protocol.

Installation

To integrate OpenLang into your OpenClaw environment, execute the following command in your terminal or command interface: clawhub install openclaw/skills/skills/andreanjos/openlang Once installed, the skill becomes available globally for your agents to utilize during sessions_spawn and sessions_send operations. No further configuration is required to begin using the protocol prefixes.

Use Cases

OpenLang is intended for back-end agent-to-agent workflows. Use it when tasking sub-agents to perform repetitive or complex operations, such as directory traversal, file batch processing, or database queries. It is ideal for orchestrating large-scale parallel tasks where the agent needs to send clear, concise instructions. Note that it should never be used for user-facing interactions (e.g., Slack, Telegram, or email), as it is optimized for machine parsing rather than human readability.

Example Prompts

1. "~openlang ~L2 ?fnd @fs {p:'src/*.ts' rec:true} !wr @fs {p:'build/log.txt' v:'Init search'}"
2. "~openlang ~L3 !git:mrg {br:'main' frc:true} >ok"
3. "~openlang ~L2 ^if {$var.status == 'ok'} {!sh:run {cmd:'npm start'}} ^el {!sh:run {cmd:'npm install'}}"

Tips & Limitations

• Default to L2: For most scenarios, the L2 (Glyph) level provides the best balance of density and reliability. Reserve L1 for complex conceptual logic and L3 for high-frequency, simple bulk data operations.
• Fallback Logic: If a receiving agent does not fully support OpenLang, the protocol is designed to be readable enough that the agent can still deduce the intent from the structural sigils.
• Variables: Always define variables with ->$name before attempting to access them as $name. Property access is supported via dot notation ($var.field).
• Safety: Always verify your command strings when using !sh or !fs scopes, as compressed commands are executed directly by the target agent.