nitro agent memory

Memory is the workspace's shared knowledge: what one session learns, every future session and every other agent starts from. Read it at session start, capture cheaply while you work, curate what proves durable. It lives locally with the repository, shared across all its git worktrees, never pushed to the remote. Always call a subcommand; bare nitro agent opens an interactive TUI and blocks the session. If commands report no workspace, run nitro agent init once from the repository root (see the sibling nitro-task skill). If nitro itself is not found, the CLI is not installed — stop and tell the user to install it: https://chillicream.com/docs/nitro/cli/installation. Do not attempt to install it yourself.

Core principles

• Memory holds what you would otherwise repeat in every prompt. Standing preferences, durable domain facts, references. Work and decisions in flight live in nitro agent tasks; messages to agents go over nitro agent mail; follow-ups and TODOs go to the tracker.
• Every agent reads everything. Reads are not filtered by actor. Writes record who saved it: pass --actor <name> on save and log, the only two memory commands that take an actor. The name is handed to you, never invented. With Nitro's hooks installed, the session-start hook states it in your context: Your Nitro actor name is "maya". Otherwise nitro agent login allocates one and prints it.
• Capture is cheap, curation is deliberate. log without ceremony; save or promote only what proved durable.
• One self-contained fact per memory. A future session has no chat context: "Prefer X over Y because Z", not "as discussed, X".
• Save only what every agent may read and what stays true. Credentials, tokens, and personal data stay out (deletion is not an erasure guarantee); volatile state -- current branch, counts, progress -- belongs in the journal or the tracker.

Recall