A personal Obsidian vault running a modified PARA method, plus the templates, queries, and Claude Code skills that make it work day-to-day.

Personal content (goals, projects, daily notes, etc.) is gitignored — what's tracked here is the system: the docs that describe how the vault is organized, the templates new pages are built from, the Dataview/Bases queries that power navigation, and the skills that automate the repetitive parts.

.claude/skills/   # Vault-local Claude Code skills (create-page, archive, audit, daily-review, place-wiki)
bases/            # .base files + Dataview/Tasks index pages
templates/        # Templater templates for each content type
wiki/system/      # Operating manual + navigation dashboards
  manual/         # Schema and workflow docs for each content type
  dashboards/     # Pinned pages: Control Panel, Current Projects, Tasks, etc.
CLAUDE.md         # Instructions Claude Code loads on every session

Everything else (archives/, areas/, daily/, goals/, habits/, projects/, and topical wiki/ subfolders) is excluded by .gitignore — placeholder .gitkeep files preserve the folder structure.

Six types, each with its own folder, frontmatter schema, and lifecycle. Full docs in wiki/system/manual/:

The PARA Method page is the entry point — it covers how the six types relate, why this deviates from canonical PARA, and how the Archives hybrid model (status flip + folder move) works.

Five vault-local Claude Code skills in .claude/skills/, documented in Skills.md:

• create-page — interactive scaffolding for a Goal/Area/Project/Habit/Wiki
• place-wiki — picks the right wiki/ subfolder, pushes back on root-dumping
• archive — atomic status flip + completion date + file move to archives/<year>/<type>/
• daily-review — morning carry-forward and evening highlights flow
• audit — vault-wide schema/drift validator

Skills delegate to the Obsidian CLI for file ops so the metadata cache stays consistent. Multi-step destructive work (archive, audit) is bundled into Python scripts with rollback.

Documented in Plugin Stack.md. The load-bearing ones:

• Templater — applies type templates on file creation
• Dataview — ## Active Projects / ## Active Goals backlink queries on Area and Goal pages
• Tasks — task filtering with useFilenameAsScheduledDate: true scoped to daily/
• Bases — .base files for the cross-type indexes in bases/
• Obsidian CLI — the API the skills drive

The structure is opinionated but not personal. To adapt it:

1. Clone into a fresh directory
2. Open as an Obsidian vault, install the community plugins (see below)
3. Copy wiki/system/dashboards/Control Panel (example).md to Control Panel.md and pin it — the original is gitignored as personal
4. Read wiki/system/manual/PARA Method.md end-to-end
5. Start using the create-page skill in Claude Code to build out your own Goals/Areas/Projects

The skills assume Claude Code with the Obsidian Terminal plugin (or any shell rooted in the vault). CLAUDE.md at the repo root is what teaches Claude the conventions.

Plugin vendored code (main.js, styles.css) is gitignored — only each plugin's data.json (config) and manifest.json (identity) are tracked. Install the plugins yourself; on first enable, Obsidian reads the tracked data.json and your settings restore automatically.

In Obsidian: Settings → Community plugins → Browse, install and enable each:

Bases is a core plugin in Obsidian 1.9+; enable under Settings → Core plugins.

Hot Reload (optional, dev-only) is gitignored entirely. Install via BRAT from https://github.com/pjeby/hot-reload only if you're authoring Obsidian plugins.