Skip to content

yogsoth-ai/context-management

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Context is not memory — it is the scaffolding that lets you build higher. Without it, every session starts from the ground floor.

📋 Context Management

Research Process Context Checkpointing Engine

Context Management persists research process and results to disk during long multi-phase research sessions. Each Phase gets its own markdown file. Each strategy completion triggers a checkpoint that appends >=500 lines of detailed process + results. The output is a comprehensive research log that survives context window compaction and serves as reference material for future sessions.


⚡ What It Does

  • 📄 context-init — creates a new context file at Phase start with plan context and metadata
  • ✍️ context-checkpoint — appends >=500 lines of research process + results at each checkpoint point

🎯 Design Philosophy

Write-Only

This repo only writes context files. It never reads them back or performs recovery. Context files are for future reference — users manually feed them to CC in new sessions when needed.

Plan-Driven Triggering

Checkpoints are triggered by the research plan. The plan embeds import context-management:context-checkpoint at appropriate points. This repo does not decide when to checkpoint — the plan decides.

CC Autonomy

Content format is CC's decision. A default semi-structured template is available but not mandatory. The only hard constraint is minimum content volume (>=500 lines per checkpoint).


📁 Repo Structure

context-management/
├── skills/
│   ├── context-init/SKILL.md
│   └── context-checkpoint/SKILL.md
├── scripts/
│   └── timestamp.py
├── README.md
└── .gitignore

🔗 Integration

Calling Flow

plan.md (generated by research-planning repo)
  │
  ├── Phase start: "import context-management:context-init"
  │     → creates context/<timestamp>-<topic>.md
  │
  ├── After strategy X: "import context-management:context-checkpoint"
  │     → appends >=500 lines to current Phase's context file
  │     → updates INDEX.md
  │
  └── Next Phase: "import context-management:context-init"
        → creates new file for new Phase

Output Location

Context files are written to context/ in the user's working directory:

<user-project>/
├── context/
│   ├── INDEX.md
│   ├── 2026-05-16-14-30-lit-survey.md
│   └── 2026-05-16-19-00-gap-analysis.md
└── ...

🛠️ Usage

Prerequisites

  • Python 3 (for timestamp script)
  • Claude Code with skill import support

Installation

claude plugin add yogsoth-ai/context-management

📄 License

Apache-2.0

About

Research Process Context Checkpointing Engine — persists research process and results to disk via >=500-line markdown checkpoints during multi-phase research sessions.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages