Works with the AI Agents you already use
duker brain
Dynamic context routing and hierarchical Obsidian vault memory for coding agents via the duker-brain MCP server.
$1
duker brain
Example session with this skill installed
Check the brain for our project conventions on asyncpg pooling and then update the database connection module.
- Read your context and instructions
- Compiled the duker brain
I have queried the project memory to retrieve our specific conventions for asyncpg.
Project Context Retrieval
| Note ID | Title | Routing Criteria |
|---|---|---|
asyncpg_pooling | asyncpg Connection Pooling | PostgreSQL connections, asyncpg pool sizing, PgBouncer transaction mode, statement cache errors |
Based on the retrieved guide, I will now update the database connection module to use min_size=5 and max_size=20 as per our internal production standards, ensuring we avoid the statement cache errors mentioned in the project memory.
import asyncio
import asyncpg
async def run():
# Applying project-specific pooling conventions from asyncpg_pooling guide
pool = await asyncpg.create_pool(
dsn='postgres://user:pass@host:port/db',
min_size=5,
max_size=20,
max_queries=50000,
max_inactive_connection_lifetime=300.0,
setup=None # Statement caching disabled per ADR-04
)
return pool
Next steps
- Run
brain_doctorto ensure no criteria conflicts were introduced if you decide to add more notes. - Verify the pool lifespan settings match the latest PgBouncer deployment by checking the
infrastructurefolder in the vault. - Call
brain_check_routingwith "how do we handle pg errors" to confirm this note remains the primary target.
Connects securely to your tools. The creator never sees your data.
About this skill
Context window limitations often lead to agents forgetting project-specific architectural decisions, library quirks, and internal coding standards. This skill integrates an Obsidian-based hierarchical memory vault via the duker-brain MCP server, enabling agents to retrieve and maintain long-term project knowledge.
What it does
- Dynamic Context Routing walks an Obsidian folder tree to fetch the specific note matching the user's intent.
- Hierarchical Memory Management organizes engineering notes into a routable structure with discriminative criteria.
- Vault Health Auditing uses
brain_doctorto detect thin criteria, over-full folders, or missing catch-all notes. - Automated Indexing ensures the router's decision model stays in sync with filesystem changes.
- Conflict Resolution manages sibling notes by rewriting criteria to ensure distinct routing paths.
How it works
- Query: Call
brain_routewith a natural language request to fetch the relevant technical guide before answering. - Explore: Use
brain_treeandbrain_searchto locate existing documentation and prevent duplicate content. - Write: Author new notes or branches using
brain_add_noteto automatically generate metadata and update the index. - Optimize: Use
brain_split_branchto keep folders under 15 items, maintaining high routing accuracy. - Verify: Run
brain_check_routingto test if specific prompts successfully land on the intended note.
Frameworks & tools
Built for the Model Context Protocol (MCP) using Node.js (>= 22). It manages markdown-based Obsidian vaults and requires the duker-brain server package.
Why this beats prompting it yourself
Standard RAG often retrieves noisy or irrelevant chunks from large files. This skill uses a tree-traversal decision model that picks the single most relevant guide based on specific, human-defined criteria, ensuring the agent follows project-specific rules over generic LLM training data.
Use cases
- Documenting team-specific PostgreSQL connection pooling and error handling patterns.
- Storing architectural decision records (ADRs) that the agent must reference before refactoring.
- Onboarding new developers by routing their setup questions to the correct environment guides.
- Managing a growing knowledge base of library-specific gotchas that generic models miss.
Known limitations
Requires Node.js 22 or higher. The router's accuracy degrades if folders contain more than 15 children. Routing requires a warm npx cache or local MCP server installation.
Requires Node.js 22+. Folders are limited to 15 children for routing reliability. Manual file edits require a brain_rebuild call to sync the index.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 day ago
Requires