9h • General
My Obsidian ↔ Claude Code Workflow
HONESTY FIRST:
Claude code wrote MOST of this post.
Okay, lets get into it.....
- Claude.md - The Instruction Manual
A markdown file at the root of my vault that Claude reads automatically every session. It teaches Claude:
- My vault's folder structure and type system (movements, recipes, tools, people, etc.)
- Frontmatter schema and naming conventions (kebab-case properties, daily note linking pattern, etc.)
- How my Obsidian plugins work (Templater, Projects, Periodic Notes)
- My rating system, tag format, file naming conventions
- Business/client project structure and SOPs
This is the foundation. Claude doesn't need to "figure out" my system each time — it already knows it.
2. Skills - Custom Behaviors
- /end-session - Saves session state when I'm done. Creates a dated session archive and updates a persistent memory file. Two-tier system: working memory (kept under 100 lines) + append-only daily archives.
- Training Coach - An always-loaded knowledge base of my training context: athlete profile, injury history, programming principles, supplement protocols, active progressions, and datestamped coaching insights. Grows over time as we (Claude and I) discuss training.
- Obsidian Nexus (the name of my vault) - Knows how to create notes that follow my vault's exact conventions (correct folder, frontmatter, wikilinks, type classification).
- Social to Typefully - Drafts and publishes social posts via MCP integration. (this deserves its own post — I practically made my own social media scheduler in Obsidian...)
3. Memory System - Continuity Across Sessions
Claude Code doesn't remember past conversations by default. I solved this with a layered memory approach:
- Persistent memory (claude-session-memory.md) - Active projects, recent decisions, action items, session context. Loaded every session. Capped at 100 lines to stay focused.
- Session archives (claude-sessions/YYYY-MM-DD session.md) - Detailed logs of each session. Multiple sessions per day append to the same file. This is the long-term history.
- Auto-memory (Claude Code's built-in) - A separate .claude/ directory that persists key patterns and lessons learned.
- Active Context files - Top of Mind.md, Body.md, and the Training Coach file are always loaded, so Claude starts each session knowing my current priorities, physical profile, and training state.
Most of this methodology is stolen from OpenClaw and how it stores memory, I just replicated it in my vault. This is tied VERY closely with my /end-session skill.
4. How a Session Actually Flows
- I open Claude Code in my vault directory
- Claude automatically reads CLAUDE.md + memory files + active context
- I work — creating notes, discussing training, writing client emails, organizing vault content
- When done, I run /end-session which extracts what happened, updates persistent memory, appends to the daily archive, and updates domain-specific files (like training coach) if relevant
- Next session picks up where I left off
----
That is pretty much it. It's been very very useful for me, and changed the way I interact with AI fundamentally. There are more hard-core business-specific uses for me (like MCP's with Notion, and doc management), but this is the core, foundational usage.
0
0 comments
My Obsidian ↔ Claude Code Workflow
powered by
skool.com/parazettel-community-4498
Home of the system combining PARA Method and Zettelkasten in PKM • Courses, discussions, workshops and more...
Suggested communities
Powered by