A Claude Code skill that automatically triggers an agent to document your technical projects in Obsidian as you work.

The skill has two execution paths:

Session start (read-only): When you open a project, the skill reads your vault note and CLAUDE.md, then

briefly orients you — current phase, status, and the next steps from last time. No writes, no agent.

Documentation run: When you ask to document, wrap up, or update notes, the skill detects project context, asks

any questions upfront, then launches the documentation agent in the background. You can keep working while your

notes are updated and synced.

The agent also performs cross-project relationship analysis each session, scanning your vault to find genuinely

related projects based on shared technologies and explicit context signals, and writes scored wiki-links into each

note's frontmatter and body automatically.

Context Detection

The skill intelligently detects project context:

1. Project Name - From git repo, directory name, or asks you
2. Area Classification - Based on file extensions and patterns (all areas counted in parallel; clear winner

wins, ties escalate to a question):

• Hardware: .ino, .pcb, .sch, platformio.ini (Arduino, embedded)
• Software: .js, .ts, .py, .go, .rs, package.json, Cargo.toml, go.mod (web, scripts, systems)
• Woodworking: .stl, .blend, .f3d, .skp, cut-list.md (CAD, shop files)
• Music Synthesis: .pd, .maxpat, .syx, .amxd, patch-notes.md (Pure Data, Max/MSP, Ableton)

1. Description - Extracts from conversation or README.md

Vault Structure

Project notes are placed into a Projects directory in your Obsidian vault. No other folders are touched. If a

Projects folder already exists, only files managed by this skill are modified. If a note with the same name already

exists, project updates are appended to it rather than overwriting existing content.