stream-terminal-output-to-agent

infrastructuretransform

Stream<StdoutChunk> -> Stream<AgentObservation>

Stream command execution stdout/stderr chunks progressively to the agentic LLM context while enforcing token-limit truncation rules.

Problem it solves

Buffering massive command outputs blocks the agent's real-time reasoning and can easily overflow LLM context window limits.

Consumes

Stream<StdoutChunk>

Emits

Stream<AgentObservation>

Distilled from 1 source

The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.