Skip to main content

​
The Agent’s Permanent Memory

Knowledge files are the agent’s permanent memory — long-term facts that persist indefinitely and are always available in context. While episodes capture what happened and consolidation compresses the timeline, knowledge files store what Wolffish knows about you, your world, and your preferences.

​
Location

~/.wolffish/workspace/brain/hippocampus/knowledge/
├── people.md
├── projects.md
├── preferences.md
├── technical.md
└── decisions.md

​
The Five Default Files

FilePurpose
people.mdPeople you interact with — names, roles, relationships, context
projects.mdActive and past projects — stack, status, key decisions
preferences.mdYour personal preferences — tools, workflows, communication style
technical.mdTechnical facts — environments, APIs, credentials locations, conventions
decisions.mdDecisions you’ve made — rationale, constraints, outcomes
These five files are created during first launch. You can add more knowledge files — the agent reads everything in the knowledge/ directory. Create books.md, health.md, or whatever categories make sense for your life.

​
How Knowledge Gets Written

Knowledge enters the system through three paths:

​
1. Automatic Promotion (LLM-driven)

During consolidation, the LLM identifies patterns that deserve permanent storage. If the same preference appears across multiple weeks, or a project keeps coming up, the consolidation process promotes it to the appropriate knowledge file.

​
2. Direct Agent Writes

When you say something like “remember that I prefer conventional commits” or “my manager’s name is Sarah,” the agent writes directly to the appropriate knowledge file during the conversation. The agents.core.md procedures instruct Wolffish to update knowledge whenever it learns a new long-term fact.

​
3. Manual Editing

Open any knowledge file in your editor and change it. The brainstem file watcher detects the change, the cortex re-indexes it, and the next conversation will use the updated information.

​
Context Priority

Knowledge files have the highest memory priority after identity files (soul.md, user.md). They are always included in context assembly — unlike episodes and consolidated files, which pass through RAS scoring. The logic is simple: if you told the agent to remember something permanently, it should always be available.

​
Example: preferences.md After a Few Weeks

# Preferences

## Development
- Package manager: pnpm (never npm or yarn)
- Commit style: conventional commits (feat:, fix:, chore:)
- Validation: Zod at all API boundaries
- Testing: Vitest for unit tests, Playwright for e2e
- Formatting: Biome over ESLint+Prettier

## Communication
- Prefers concise answers with code examples over lengthy explanations
- Likes bullet points over paragraphs for technical content
- Wants to be warned before destructive operations, even if flagged safe

## Workflow
- Morning sessions (08:00–12:00) for deep work — avoid scheduling interruptions
- Prefers creating a new branch per feature, never commits directly to main
- Likes PRs reviewed before merge, even on solo projects

## Tools
- Editor: VS Code with Vim keybindings
- Terminal: Warp
- Browser: Arc for development, Chrome for testing
- Notes: Obsidian with daily notes

​
Example: people.md

# People

## Sarah (Manager)
- Engineering manager at current company
- Prefers async communication over meetings
- Reviews PRs on Tuesdays and Thursdays

## Omar (Co-founder)
- Working on the billing service together
- Handles infrastructure and deployment
- Timezone: UTC+3 (Riyadh)

## Alex (Design)
- UI/UX designer on the product team
- Shares Figma links for review — always check mobile breakpoints

​
Editing Guidelines

Just tell Wolffish during conversation:
  • “Remember that my API key for Stripe is in 1Password”
  • “Note that we decided to use PostgreSQL over MySQL for the new service”
  • “My flight to Riyadh is on June 3rd”
The agent will write to the appropriate knowledge file.
Knowledge files work best when they’re concise. Each entry should be a fact, not a story. If you find a knowledge file growing past 100 lines, consider splitting it into multiple topic files.

​
Custom Knowledge Files

Create any .md file in the knowledge/ directory and it becomes part of the agent’s permanent memory: 
# Create a new knowledge file
touch ~/.wolffish/workspace/brain/hippocampus/knowledge/health.md
The brainstem file watcher will detect it, the cortex will index it, and prefrontal will include it in context assembly with the same priority as the default five files.