bigredt/legible

How a coding agent presents its tool calls and their results so a human can follow the session without reading raw output. Shell commands are written as commented narration with labeled, filtered output, and every tool call (or batch of reads) is followed by a "↳ Execution summary" line saying what came back, what changed and what it means. It changes presentation only, not what the agent does. Use this for every shell command, file edit or write, external service or MCP call, web fetch, subagent, and batch of file reads or searches, for the whole session, including short status checks and routine steps, even when nobody has mentioned it.