Session-start briefing

When a session opens, before your first prompt, the Neuronz.ai plugin gives your agent a short briefing for the profile of the directory you're in. Auto-recall takes over for each prompt after that.

What the briefing contains

• A short guide to using memory: where each kind of information belongs (fact, rule, persona trait, knowledge doc, past work) and which source wins when two disagree (see below).
• The rules that apply: how many approved rules this profile has, their names, and the call that fetches their text. The text itself isn't included, so the agent reads the rules before it acts. Proposed rules aren't listed until you approve them.
• The topic you're working in, if the session has one open: its name and the call that loads its brief. See topics.
• Your profile: up to 25 of your own persona traits, in full. Later in a long session they are repeated every 20 prompts and right after the agent compacts its context.
• A count of stored facts: how many verified facts the profile holds, a reminder that none of them are in the agent's context, and the instruction to search before answering. Persona traits, unverified claims and plans aren't counted.

In the same step the plugin also makes this profile's skills and commands available to your agent.

When space runs out

Coding agents cut off an oversized session-start message, so the briefing has a size limit. Each section keeps a minimum short form: a single line saying what was left out and how to fetch it. A section can shrink to that line but never disappears. The remaining space goes first to your rules. Sizes are counted in characters, so a briefing in French or Japanese gets as much room as one in English.

When a fact and a document disagree

The briefing tells the agent how to decide when a stored fact and a checked-in document such as CLAUDE.md disagree:

• How things should be done (conventions, standards, intent): your reviewed docs and approved rules win over a stored fact.
• How things are right now (versions, paths, what's wired to what): the code wins. Stored facts and knowledge are usually more up to date than a checked-in .md file.

The agent fixes whichever side is wrong: it corrects a stale fact, and proposes a change to a stale document rather than rewriting it silently.

Using it

The briefing is automatic once the plugin is installed; there's nothing to call. It opens with a line naming the profile, for example:

Persistent memory is active for profile "billing-service".

Curate facts, your persona and your rules in the dashboard; the next session starts with the updated picture.

Related

• Rules — the rules the briefing lists, and how rules reach the agent as you work.
• Personas & the social map — the traits included in the briefing.
• Auto-recall — what the agent receives with each prompt.
• The plugin & the automatic loop — what runs, and when.