Skip to content

TAIKER656/claude-memory-discipline

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 

Repository files navigation

Claude Memory Discipline

Stop forgetting your own commitments. A 3-layer memory system for Claude Code power users running multiple projects.

License: MIT Claude Code

The Problem

Claude Code's auto-memory is powerful but has three gaps that bite power users:

  1. Per-cwd loading — Memory loads per project directory. Run a session in ~/Desktop/project-a/, critical commitments from project-b/ become invisible.
  2. MEMORY.md truncation — Auto-loaded index has a ~24KB ceiling. Power users exceed this in 2-3 months. Truncated content = silently invisible.
  3. No "what's due today" radar — Session logs are chronological prose. No structured view of TODAY items / OVERDUE / approaching commitments.

Result: You write plans to memory. Future-Claude doesn't proactively surface them. User has to remember to ask. If user forgets → critical commitment slips.

The Solution — 3-layer system

Layer 1 — Passive auto-load files

  • calendar_active_dates.md — Central dates radar (TODAY/TOMORROW/OVERDUE)
  • active_pipeline_overview.md — Central contacts dashboard (who's waiting reply)

Layer 2 — Active workflow skill

  • pulse-check — Auto-runs on greeting trigger ("siema"/"hey"/"what's up"/"co robimy"). Reads calendar + pipeline + recent session logs. Outputs structured TODAY/TOMORROW/OVERDUE summary BEFORE answering user's question.

Layer 3 — System discipline rules

  • feedback_memory_discipline.md — Non-negotiable rules for future-Claude:
    • ALWAYS pulse-check on greeting
    • Date mentioned in convo = auto-save to calendar
    • Interaction (DM/email/comment) = auto-update pipeline
    • End-of-session ALWAYS update calendar + pipeline (not just session log)
    • Weekly review Pn morning trigger
    • MEMORY.md size watch (cleanup gdy >22KB)

Quick Start

  1. Clone this repo:

    git clone https://github.com/TAIKER656/claude-memory-discipline.git
    cd claude-memory-discipline
  2. Install pulse-check skill:

    cp -r pulse-check ~/.claude/skills/
  3. Copy templates to your memory dir:

    # Find your project memory dir (typically ~/.claude/projects/{slug}/memory/)
    MEMDIR=~/.claude/projects/-Users-{your-username}/memory
    cp templates/calendar_active_dates.template.md $MEMDIR/calendar_active_dates.md
    cp templates/active_pipeline_overview.template.md $MEMDIR/active_pipeline_overview.md
    cp templates/feedback_memory_discipline.template.md $MEMDIR/feedback_memory_discipline.md
  4. Customize templates — fill in your actual dates + contacts (see docs/customization.md)

  5. Test — start new Claude Code session, type "hey" or "siema" → pulse-check should auto-run

Why this works

Built and field-tested by a solo founder running 4 simultaneous projects (compliance audit consulting + 3 product builds + weekly content production). The system handles:

  • Calendar with 20+ critical dates spanning 12 months
  • Pipeline with 15+ warm contacts across multiple platforms
  • 30+ session logs accumulated over 3 months
  • Memory total ~150KB across 80+ files

Without this system: regularly missed Yaseen demo dates, Adeyinka call deadlines, LinkedIn cool-down ends. With this system: zero missed commitments over [X tracking period].

Repository contents

claude-memory-discipline/
├── README.md                                        # This file
├── LICENSE (MIT)
├── pulse-check/
│   └── SKILL.md                                    # Universal pulse-check skill
├── templates/
│   ├── calendar_active_dates.template.md          # Dates radar template
│   ├── active_pipeline_overview.template.md       # Contacts dashboard template
│   └── feedback_memory_discipline.template.md     # Discipline rules template
├── examples/
│   └── anonymized-snapshot.md                     # Real-world usage example
└── docs/
    ├── installation.md                             # Step-by-step setup
    ├── customization.md                            # How to adapt for your workflow
    └── why-this-exists.md                          # Origin story + design decisions

Author

Built by Piotr Reder — Founder at eucomplyhub.com (EU AI Act audit consulting for mid-market SaaS).

Origin: 15.05.2026 — User feedback "you save plans to memory but then forget them. If I forget too, critical commitment slips." → 3-layer system designed in response.

Contributing

PRs welcome. Open issues for:

  • Bug reports
  • New template ideas (e.g., research-pipeline overview, content-calendar)
  • Documentation improvements
  • Translation (current: English)

License

MIT — see LICENSE.

See also


⭐ If this saved you from forgetting an important commitment, please star the repo.

About

3-layer memory system for Claude Code power users running multiple projects

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors