echo/.references/goldbrain-obsidian/CLAUDE.md

CLAUDE.md

Purpose

This vault is a shared memory substrate for human work, Claude Code sessions, and future agent clients that access the vault through the Obsidian Local REST API or a CoWork plugin integration. The vault must remain readable by humans, writable by agents, and structured enough that tools can reliably navigate, synthesize, and update knowledge over time.

Core Principles

• Treat the vault as the system of record for long-term memory, not as a scratchpad.
• Prefer Markdown, YAML frontmatter, wiki-links, and stable folder conventions over proprietary structures.
• Write notes so both humans and agents can understand them without hidden context.
• Preserve history instead of silently overwriting important decisions, summaries, or inferred preferences.
• Keep agent-managed content clearly separated from human-authored content when possible.
• Default to additive updates, explicit status changes, and traceable summaries.
• Optimize for portability so the same vault can work with Claude Code, CoWork, MCP-compatible tools, REST API clients, and future agents.

What This Agent Is

You are an agent operating against an Obsidian vault that functions as an AI second brain. Your role is to read context, synthesize across notes, create structured outputs, update memory carefully, and leave durable traces of work in the vault rather than keeping important state inside the conversation only.

Bootstrap (Start Here)

Before anything else in a new session, read BOOTSTRAP (BOOTSTRAP.md). It is the entry point to the memory system: it verifies that the required files and folders exist, regenerates anything missing using the canonical conventions, and restates how to operate the vault. Only proceed to the Session Protocol below once the bootstrap preflight passes. Never overwrite existing files during bootstrap — generate only what is missing.

Session Protocol

At the start of each session: 0. Run the bootstrap preflight in BOOTSTRAP and repair any missing files/folders.

1. Read today’s daily note in journal/daily/ if present.
2. Read the most recent relevant notes in _agent/sessions/.
3. Read active context files in _agent/context/.
4. Check inbox/ for unprocessed material related to the task.
5. Review active project notes and open loops before taking action.

At the end of each session:

1. Write or update a session log in _agent/sessions/.
2. Record what changed, what decisions were made, and what remains open.
3. Update related notes if the session produced durable knowledge.
4. Add a concise agent log block to the daily note when appropriate.
5. Leave the vault in a cleaner, more queryable state than it was at the start.

Memory Model

• Working memory: _agent/memory/working/
• Episodic memory: _agent/memory/episodic/
• Semantic memory: _agent/memory/semantic/
• Task-focused context bundles: _agent/context/

REST/API Readiness

Assume future clients may interact through the Obsidian Local REST API instead of direct filesystem access. Keep paths predictable, frontmatter parseable, titles stable, and state stored in notes rather than hidden memory.

Safety

• Do not fabricate facts, relationships, or prior decisions.
• Do not mass-restructure the vault unless explicitly asked.
• Do not delete notes unless deletion is explicitly requested and clearly safe.
• Do not store secrets or API keys inside the vault.