Agent Claude Code
Use whenever the main agent needs to recall past work — episodes, prior sessions, decisions, or past artifacts — from inside .xo/. Keeps heavy reads off the main thread and returns only distilled, relevant context. Invoke when the user references prior work ("continue", "the bug from yesterday", "what we discussed")…
Agent Claude Code
Use at the end of every session — when the user signals stop ("done", "wrap up", "good for today") or a natural close is detected. Runs the full session-close ritual: writes session.md, distills semantic memory, promotes artifacts, updates WORKSPACE.md, runs cheap compression, wipes working memory. Always dispatch…
Agent Claude Code
Use when the main agent is about to perform a recurring workflow (deploy, release, migration, code review, debugging recipe) and may have a procedural skill stored in .xo/memory/procedural/ or .xo/skills/. Returns the matching skill verbatim or a clear "no skill found".
Instructions file CodexOpenCode
AGENTS.md instructions for dwivedi-ai/ultimate-work, covering agents.md, stable prefix (auto-loaded), boot ritual, during-session rules and memory writing discipline.
Instructions file
Claude Code instructions for dwivedi-ai/ultimate-work.