Generated from modular-pfc commit
9ef7cd0on 2026-05-20.
A modular file-based productivity system designed for people with ADHD, autism, alexithymia, or other prefrontal-cortex / executive-function challenges. Optimized for use with Claude Code (also works with Codex and any agent that reads AGENTS.md).
PFC = Prefrontal Cortex. The system externalizes everything an executive function would normally hold internally: tasks, focus, habits, day ratings, projects, areas of life, values, visions, insights. The git repo is canonical; chat-based agents (Claude Code, etc.) read and write to it.
The bias is toward less surface for new users on day one — you don't need to learn it all to get value. There are 30+ slash-command skills, but most users only touch ~7 in their first week.
This is the most important thing to know on day one. PFC is a starter kit, not a fixed product. The shipped skills, schemas, area folders, and rules are defaults — every one of them is meant to be edited.
Examples of things you can ask your agent to do:
- "Edit
pfc-morning-checkinto also ask about water intake." - "Add a new skill that pulls weather and tells me if it's a good outdoor-walking day."
- "Delete the
legalarea — I don't track regulatory work." - "Change the daily-focus model from 2+1 to 3+0 — I don't want the bonus slot."
- "Add a new field
caffeine_mgto the day-tracker schema and prompt me for it during evening-checkin."
The system is built to be modified, not just consumed. When something doesn't fit your life, change it. Don't work around it. Run /pfc-onboarding meta/system-is-editable for the longer pitch.
For the full philosophy of how the system thinks, see AGENTS.md.
- Use this template (top-right green button on github.com).
- Clone your fork locally.
- Open in Claude Code (
claudeCLI) or any agent that readsAGENTS.md. - Run
/pfc-onboarding. The skill walks you through the system one feature at a time, paced to weeks/months instead of minutes.
That's it for Tier 1. The system works with markdown alone — no Python venv, no cron, no API tokens.
| Tier | What you get | Setup time | Doc |
|---|---|---|---|
| 1 | Markdown system + slash commands | 5 min | This README |
| 2 — Local automations | Auto-fetched sleep/AZM, scheduled jobs | 30 min | docs/install/local-automations.md |
| 2 — Connectors | Google Calendar + Gmail integration | 15 min | docs/install/connectors.md |
| 2 — Trello dashboard (recommended) | Visual phone-glance view + mobile capture widget | 20 min | docs/install/trello.md |
| 3 — VPS | 24/7 background automation | 30 min | docs/install/vps.md |
| 3 — Remote access | Drive the VPS from your phone (Termius + tmux + Claude Code Remote Control) | 20 min | docs/install/remote-access.md |
Trello is highly recommended for the visualization benefit — externalizing the system into a glanceable visual layer is especially valuable for the target audience. Power-Ups are free.
The system follows a VAVPAH model: Values → Areas → Visions → Projects → Actions → Habits.
| Folder | What it holds |
|---|---|
0-me/ |
Who you are: profile, working-with-me, schedule patterns |
1-values/ |
Your values list (start with the menu of 30 in values.md) |
2-areas/ |
Areas of life — career, health, finances, relationships, etc. |
3-visions/ |
Long-term direction tied to an area (no deadlines) |
4-projects/ |
Short-term projects (1-3 months, max 3 active) |
5-actions/ |
Tasks, daily focus picks, event log |
6-habits/ |
Daily + monthly habit tracking |
data/ |
Day tracking, insights, hypotheses, findings, supplements |
config/ |
Schemas + integration config |
docs/ |
System documentation |
automations/ |
Python scripts for auto-fetched data + systemd units |
.agents/skills/ |
Skill definitions (read by Codex and most other agents) |
.claude/skills/ |
Symlink mirror for Claude Code |
The canonical instruction file is AGENTS.md — read by Codex and most other agents that follow the AGENTS.md convention. CLAUDE.md is a one-line @AGENTS.md import for Claude Code's own loader. The skills work with any agent that has a Skill-equivalent invocation pattern; the slash commands work natively in Claude Code.
The PFC template keeps evolving — new skills, scrubs, install-doc improvements, refined defaults. Your fork is its own thing; you don't have to stay in lockstep. If you want to pull selected upstream updates into your fork later, the /pfc-pull-template-updates skill walks each upstream commit one at a time and lets you accept or reject per commit.
Three things to know before you pull updates:
- The maintainer tailors the template to his own workflows. Some updates will fit you; some won't. The system was always meant to be edited — the same applies to incoming changes. Pick the refinements you actually want, drop the rest.
- Upstream changes can clash with your modifications. Heavily-customized skills, areas, or schemas may produce merge conflicts. Your modifications take priority — pull updates only when you genuinely want the upstream behavior.
- Pulling is opt-in. A fork that never re-syncs with upstream is entirely valid. Pull when you're curious or when a specific upstream change interests you, not because you "should" stay current.
If a default doesn't fit your life, modify it locally rather than fighting upstream commits forever. That's the spirit of the original meta/system-is-editable lesson — and it applies just as much to upstream-tracking as it does to day-one edits.
MIT. See LICENSE for the full text.