Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.wolffi.sh/llms.txt

Use this file to discover all available pages before exploring further.

​
Workspace Architecture

Everything Wolffish knows, remembers, and can do lives in ~/.wolffish/workspace/. This is your agent’s brain on disk — readable, editable, and git-versionable.

​
Directory Structure

~/.wolffish/workspace/
└── brain/
    ├── identity/
    │   ├── soul.md                    # Agent personality
    │   └── user.md                    # User profile
    ├── prefrontal/
    │   ├── agents.md                  # Agent procedures + tool usage rules
    │   └── .debug/                    # Context snapshots (auto-created)
    ├── cerebellum/                    # Capabilities
    │   ├── shell/                     # Shell commands (SKILL.md + plugin/)
    │   ├── filesystem/                # File operations (SKILL.md + plugin/)
    │   ├── git/                       # Git operations (SKILL.md only)
    │   └── introspect/                # Self-awareness (SKILL.md + plugin/)
    ├── hippocampus/
    │   ├── episodes/YYYY-MM-DD.md     # Daily conversation logs
    │   ├── consolidated/YYYY-WNN.md   # Weekly summaries
    │   └── knowledge/                 # Long-term facts (topic files)
    ├── motor/tasks/TASK-{id}.md       # Task execution logs
    ├── basalganglia/YYYY-MM-DD.md     # Daily feedback logs
    ├── brainstem/heartbeat.md         # Cron schedule definitions
    ├── corpus/YYYY-MM-DD.log.md       # Daily event logs
    ├── conversations/                 # Chat history
    ├── cortex.db                      # SQLite FTS5 index (disposable)
    └── config.json                    # App configuration

​
Key Principles

Markdown is truth, cortex.db is cache. The SQLite database is fully disposable. Delete it and Wolffish rebuilds it from the markdown files on next launch. Never edit cortex.db directly — edit the markdown source files instead. Daily files over growing files. Episodes, feedback, and event logs are split by day (YYYY-MM-DD.md). This keeps individual files small, makes them easy to browse, and allows the brainstem to clean up old logs automatically. Capabilities are self-contained. Each folder under brain/cerebellum/ is a complete capability. You can delete any of them without affecting the core system. You can add new ones by dropping in a folder.

​
Inspecting the Workspace

You can open ~/.wolffish/workspace/ in VS Code or any text editor to see your agent’s entire state. Some useful things to inspect:
What You Want to KnowWhere to Look
What context was sent to the LLMbrain/prefrontal/.debug/context-*.md
What the agent remembers about todaybrain/hippocampus/episodes/YYYY-MM-DD.md
What tools were called and what happenedbrain/motor/tasks/TASK-*.md
Whether the agent is learning your preferencesbrain/basalganglia/YYYY-MM-DD.md
Every event that firedbrain/corpus/YYYY-MM-DD.log.md

​
Version Control

You can git init the workspace directory to version-control your agent’s brain. This lets you track changes to personality, review memory evolution, roll back configuration changes, and share your agent setup with others. 
cd ~/.wolffish/workspace
git init
echo "brain/cortex.db" >> .gitignore
echo "brain/prefrontal/.debug/" >> .gitignore
git add .
git commit -m "Initial brain state"