How does Context Keeper track changes during a session?

It uses integrated shell hooks (PostToolUse and Stop) to monitor file modifications and generates a mandatory action list that prompts the agent to update documentation before the session ends.

Can I use Context Keeper with an existing codebase?

Yes. You can run the included scan_project.py script on any existing project to detect the tech stack, infer conventions, and generate an initial USERAGENTS.md project index.

Which programming languages are supported?

Context Keeper includes pre-defined conventions for TypeScript, React, Go, and Python, but its framework is language-agnostic and can be used with any tech stack.

What happens if I forget to update the documentation?

In strict mode, the session-end hook will block the agent from finishing the task until the modified files' headers and technical info files have been updated.

Does it clutter my repository with documentation files?

No. Directory-level documentation (TECH_INFO.md) is created on-demand only for the folders you actually modify, and these files are automatically added to .gitignore to keep your remote repository clean.

Context Keeper

Name: Context Keeper
Author: royisme

byroyisme

0•

学习与文档

Maintains persistent project context through an automated, multi-layered documentation system that keeps AI agents synced with your codebase.

Context Keeper is a specialized tool for Claude Code designed to solve the problem of context drift in complex software projects. It implements a rigorous three-tier documentation strategy—spanning project-level indices, directory-level technical summaries, and file-level headers—to ensure that AI agents always operate with accurate, up-to-date knowledge. By utilizing event-driven hooks and on-demand generation, it minimizes documentation overhead while enforcing synchronization after every code modification. This makes it an essential skill for large-scale refactoring, onboarding to legacy codebases, and long-term project maintenance where architectural clarity is paramount.

主要功能

01On-demand directory documentation to prevent unnecessary file clutter

020 GitHub stars

03Built-in pre-modification checklists to ensure architectural alignment

04Session-end enforcement hooks that block completion until docs are updated

05Automated tech stack detection and coding convention inference

06Three-level documentation hierarchy (Project, Directory, and File levels)

使用场景

01Initializing context for legacy codebases to help AI understand complex architecture

02Enforcing strict coding standards across diverse stacks like React, Go, and Python

03Maintaining real-time technical change logs and dependency maps in large teams

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add royisme/agent-skills context-keeper

For use in Claude.ai and ChatGPT

Download Skill