Second Brain Skill: A Reusable Read-and-Update Workflow

A second brain skill is a reusable instruction that tells an AI assistant how to read your project record at the start of a session and how to update it before stopping. Without this skill, each session starts from zero and nothing the assistant learns persists. Tools like fenbs store the work record the skill reads and updates: projects, tasks, decisions, lessons and where you left off.

The value of a second brain comes from keeping it current. The skill is the mechanism that makes "current" automatic rather than a manual chore. Write it once, call it at the start and end of each session, and the record stays useful.

What a second brain skill does

The skill has two phases. The read phase: the assistant reads the current state of your project before starting work. The update phase: it saves a progress note and a handoff before stopping. Both phases are necessary. A skill that only reads improves input quality but adds no value to the next session. A skill that only saves captures output but not the context the next session needs to make sense of it.

A second brain for AI differs from a personal knowledge management system in that it holds the work, not facts about you. Projects, tasks, decisions and progress belong in the AI second brain; your reading notes, interests and personal archives belong elsewhere. For the broader concept of what an AI second brain is, and how it compares to a general AI brain, those posts cover the background.

The structure of a second brain skill

A second brain skill has four parts:

• Trigger: when the skill runs. Typically at the start and end of every work session, or before and after each task.
• Sources: what to read in the read phase. Rules and decisions first, then current task status, then where you left off.
• Review gate: before acting, summarize what was read and confirm the next step matches the current rules. This prevents acting on stale information.
• Update: what to save in the update phase. Progress on each task touched, the updated handoff note, and any new decision or lesson.

The review gate is the part most often skipped. Without it, the assistant reads the context and then proceeds without checking whether what it is about to do is consistent with what it just read. A one-sentence check, "the next step is X, which is consistent with rule Y and the current status", catches most misalignments before they cause rework.

A copyable template

Save the following in your AI tool's system configuration, in a Skills entry in fenbs, or in a CLAUDE.md or AGENTS.md file for Claude and Claude Code:

START OF SESSION
1. Read project AI context: Decisions and rules first, then Where we left off, In progress, Waiting on you.
2. State the current task and next step before doing any work.
3. If a rule is unclear or the next step conflicts with a rule, ask before proceeding.

END OF SESSION
1. For each task touched: update progress (what was done, what was checked, what remains).
2. Update Where we left off: next starting point, anything waiting on a person, anything still running.
3. If a decision was reached or a mistake was made, record it in Decisions and rules or Lessons.

This template works with Claude over MCP connected to fenbs. For clients that use Claude skills, see Claude skills examples and how to create a Claude skill for the setup steps.

Review and update gates

A review gate at the start of the session and an update gate at the end are the two safeguards that prevent the record from drifting out of date. The start gate ensures the assistant acts on current rules, not stale ones. The end gate ensures the record reflects what actually happened, not what was planned.

Build the gates into the skill explicitly. "State the current task and next step before doing any work" is a start gate. "For each task touched, update progress before stopping" is an end gate. Making them explicit rather than implied means the assistant is less likely to skip them when the session gets busy.

History in fenbs records every change with who made it, so if a session updates the wrong task or records an incorrect status, you can see what changed and correct it. That audit trail makes the record trustworthy enough to act on.

fenbs does not automatically run the skill; your assistant calls it at the start and end of each session on your instruction. The advantage is that the skill and the record it reads are in the same place, visible in History, and available to every supported AI tool without copying anything between apps.

How to do this with fenbs

1. Create a free account and add your project.
2. In Skills and prompts, save the second brain skill template above, adapted for your project. Give it a short name.
3. In Decisions and rules, add the standing rules for this project. Keep each one short with the reason it was chosen.
4. Connect your AI tool from Integrations. At the start of each session, ask your assistant to call the skill by name.
5. At the end of each session, ask it to call the end-of-session steps. Check that Where we left off was updated before closing.

For projects where several people work with different AI tools, the skill in fenbs ensures they all read and update the same record. A designer using Claude and a developer using Claude Code both start from the same project state when they call the same named skill. The record stays consistent without manual coordination.