- Home
- Skills
- Agents & Orchestration
- Agent Brain Skill
Agent Brain Skill
A local Markdown-based memory vault for tracking project context, technical lessons/errors, and handoffs across AI sessions.
Free
Works with the AI Agents you already use
Agent Brain Skill
Example session with this skill installed
We just finished the Auth0 integration. Record that we switched to the Authorization Code Flow with PKCE because the implicit flow was causing token renewal issues in the SPA.
- Read your context and instructions
- Compiled the agent brain
I have recorded the decision and the technical reasoning in the project memory. A new pending contribution has been created for the curator to merge into the project overview.
Memory Capture Result
| Field | Value |
|---|---|
| Project | auth-service |
| Kind | session |
| Status | Captured |
| Contribution ID | contrib_20231027_PKCE_auth |
| Path | .brain/inbox/auth-service/session_20231027_1430.md |
| Evidence | observed (Integrated Auth0 SDK and verified PKCE handshake) |
Summary of Captured Entry
Moved from Implicit Flow to Authorization Code Flow with PKCE.
- Reason: Addressed token renewal failures observed in the Single Page Application.
- Evidence: Verified successful token exchange and silent refresh using the new flow.
- Next Action: Update the client-side configuration documentation to reflect the new
redirect_urirequirements.
Next steps
- Run
python scripts/brain.py --root ".brain" statusto see all pending contributions. - Use the
publishcommand to merge this intoprojects/auth-service/overview.md. - Check
mistakes.mdto see if a generic lesson regarding implicit flow should be added.
Connects securely to your tools. The creator never sees your data.
About this skill
Development projects often lose critical context between chat sessions, leading to repeated mistakes and redundant onboarding. This skill implements a durable Markdown-based "memory" system that tracks project states, user preferences, and technical lessons across different agents and harnesses.
What it does
- Persistent memory maintains a local Markdown vault to store project overviews, practices, and mistakes.
- Context retrieval searches for project-specific pitfalls and historical decisions before starting new tasks.
- Lesson capture records technical failures with their specific symptoms, causes, and verified prevention steps.
- State tracking documents milestones, evidence, and next actions to ensure smooth handoffs between sessions.
- Memory curation merges atomic contributions into canonical project files while preserving source provenance.
How it works
- Locate the brain via
AGENT_BRAIN_ROOTorbrain.local.jsonand verify the environment has Python 3.10+. - Retrieve project context using
brain.py searchandreadto check for relevant mistakes or practices. - Capture new observations or milestones using the
capturecommand to create a pending contribution. - Curate the vault by merging pending notes into the
overview.mdormistakes.mdusing thepublishcommand.
Frameworks & tools
This tool uses a Python 3.10+ CLI helper (scripts/brain.py) to manage a local filesystem vault. It stores data in UTF-8 Markdown files and manages state via a structured JSON configuration.
Why this beats prompting it yourself
Relying on a system prompt to remember past sessions is unreliable and consumes token window. This skill provides a structured CLI for atomic updates and deduplication, ensuring that "lessons learned" from a bug in one chat are actually available to the agent in the next.
Use cases
- Preventing the agent from suggesting a library or approach that previously failed in the project.
- Resuming a complex multi-step migration after the context window has cleared.
- Standardizing coding practices across multiple agents working on the same repository.
- Managing handoffs between a "researcher" agent and a "coder" agent using shared evidence.
Known limitations
Requires a local Python environment and filesystem access. It does not provide real-time cloud synchronization or an automated agent scheduler.
Requires Python 3.10+. Manual curation is needed to merge contributions into canonical notes. No built-in cloud sync or multi-user locking beyond basic atomic file writes.
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
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 7 days ago
- Free with an Agensi account
Requires