Record what was just worked on into the MEMORY LOG — an entry file in .claude/memory/entries/, one index line in .claude/MEMORY.md, and any durable trap into .claude/memory/gotchas.md — so the next agent starts with that context instead of rediscovering it. Use at the end of any task that changed code or behavior, or…
Shape every reply to the user into one fixed layout — ==== YOU ASKED ==== (the refined prompt prompt-daddy logged, verbatim), ==== OVERVIEW ==== (what happened, a few plain sentences), ==== ACTION REQUIRED ==== (present if and only if the user must do something before the work is complete: numbered steps, exact…
Keep TERMS.md — nebula's shared glossary of ALL-CAPS canonical names for its features, panels, keys, CLI commands, hook routes, daemon mechanisms, statuses and dev workflows — true after a task. Use at the end of every task, right after nebula-memory. Every run detects the vocabulary the task surfaced and runs…
Before starting any new task, rewrite the user's prompt once into the best fully specified version of it — the ambiguous words closed, the unstated "keep X as-is" named, the why and the evidence in place, the user's aliases replaced by the ALL-CAPS TERMS from TERMS.md — asking the user only for the context the work…
Cut a nebula release — verify the tree is green, commit, bump the workspace version, tag, push, and replace the auto-generated GitHub release notes with a real changelog. Use when the user says "do a release", "cut a release", "release this", "ship it", "commit and push and release", or asks for a new version. Also…