The Problem

People who use Claude Code often start each session with a blank slate. They lose context, repeat work, and struggle to maintain a coherent knowledge graph. Ars Contexta replaces the “blank” with a self‑generating second brain that persists across sessions, reducing cognitive load and eliminating manual scaffolding.

What This Does

  • Generates a vault of plain Markdown notes linked by wiki syntax, forming a traversable graph.
  • Builds a processing pipeline of skills that extract insights, find connections, and update old notes, all driven by research claims stored in the methodology/ folder.
  • Installs hooks (defined in hooks/hooks.json) that enforce structure, monitor maintenance, and auto‑commit changes, keeping the vault healthy without manual intervention.
  • Creates Maps of Content (MOCs) at hub, domain, and topic levels, and produces a user manual from generators/claude-md.md.

How It Is Wired

  • Entry Point – /.claude-plugin/plugin.json registers the plugin’s commands (/arscontexta:setup, /arscontexta:help, etc.).
  • Command Dispatch – When /arscontexta:setup is invoked, the plugin runtime calls the shell script hooks/scripts/session-orient.sh.
  • Session Setup – session-orient.sh orchestrates the 6‑phase setup flow described in the README, invoking hooks/scripts/auto-commit.sh to capture the initial vault state.
  • Generation – The core generation logic resides in generators/claude-md.md and the feature‑specific Markdown files under generators/features/. These templates are processed by the plugin’s internal renderer, emitting Markdown into the notes/, self/, and ops/ directories.
  • Hook Execution – hooks/hooks.json maps file‑system events to scripts (auto-commit.sh, write-validate.sh, vaultguard.sh). When a note is written, the hook system triggers write-validate.sh to ensure schema compliance and then auto-commit.sh to push the changes to the local repo.
  • Side Effects – All effects are confined to the local filesystem: creation of Markdown files, updates to .git metadata, and optional local shell commands. No external network calls are made.
  • Blast Radius – The most widely used component is hooks/hooks.json; altering it changes every write operation. The generator templates have a smaller blast radius, affecting only the vault structure.

How To Use It

  1. Clone ``bash git clone https://github.com/moses-y/arscontexta cd arscontexta ``
  2. Install the plugin in Claude Code (as per README) ``bash /plugin marketplace add agenticnotetaking/arscontexta /plugin install arscontexta@agenticnotetaking ``
  3. Run the setup wizard ``bash /arscontexta:setup ` Answer the conversational prompts; the plugin writes the vault to notes/, self/, and ops/`.
  4. Activate hooks – restart Claude Code to enable the file‑system hooks defined in hooks/hooks.json.
  5. Explore – run /arscontexta:help to see available commands and /arscontexta:ask <query> to query the generated knowledge graph.

Real‑World Use

A consultant working with a legal team can use Ars Contexta to capture case notes, automatically link precedent chains, and maintain a live wiki of regulatory changes.

# After capturing a new case note
/arscontexta:ask "What precedent links to this case?"
# The plugin returns a Markdown summary with wiki links to previously stored precedent notes.

Code Health & Issues

  • Missing CI/CD – No .github/ or CI configuration.
  • Test Coverage – 18 test files exist, but no automated test runner is defined.
  • License – MIT license present.
  • Dependencies – None declared; the repo relies solely on shell scripts and Markdown, simplifying maintenance.

The Bottom Line

Ars Contexta delivers a lightweight, conversationally driven second brain without external databases or templating. It is ideal for teams that value privacy, want zero‑config setups, and can work with a shell‑based hook system. The lack of CI and automated testing is a risk for production use, but the clear folder layout and self‑contained scripts make it straightforward to audit and extend.