obsidian-mind: a persistent memory vault for AI coding agents to track professional context and performance evidence

What it solves

AI coding agents (like Claude Code) typically start every session from zero, losing context on goals, team dynamics, and past decisions. Obsidian Mind provides a persistent, structured memory system for these agents, allowing them to maintain a long-term knowledge graph of a user's work, projects, and professional growth.

How it works

The project is an Obsidian vault template designed to be used as the agent's working directory. It uses a combination of procedural hooks and agent-driven content management:

• Lifecycle Hooks: TypeScript scripts that automatically inject context (like "North Star" goals) at session start, classify user prompts in real-time, and validate markdown files after they are written.
• Vault-First Memory: Durable knowledge is stored in git-tracked markdown files. The agent manages these notes, linking projects to people and decisions.
• Tiered Loading: To save tokens, it uses a lightweight initial context injection and relies on semantic search (via the optional QMD tool) to pull specific information on demand.
• Agent Integration: It integrates with Claude Code, Codex CLI, and Gemini CLI via configuration files and hooks.

Who it’s for

Software engineers and developers who use AI coding agents and want their AI assistant to remember their professional context, track their achievements for performance reviews, and manage their project documentation automatically.

Highlights

• Automated Context Injection: Automatically loads active projects and tasks at the start of every session.
• Semanic Search: Optional integration with QMD for high-accuracy retrieval of vault content.
• Performance Tracking: A built-in system to link work notes to competency frameworks, automatically generating evidence for performance reviews.
• Specialized Subagents: Dedicated agents for tasks like "brag-spotting" (finding uncaptured wins) and "slack-archaeology" (reconstructing incidents from Slack).
• Agent Agnostic: Works across multiple CLI-based AI agents using a shared set of markdown conventions and scripts.