- Home
- Skills
- Agents & Orchestration
- Vault Recall: Load Project Memory Into Any Agent
Works with the AI Agents you already use
Vault Recall: Load Project Memory Into Any Agent
Starts every session with a short recap of past decisions, solved problems and open TODOs from your Obsidian vault.
Free
Vault Recall: Load Project Memory Into Any Agent
Example session with this skill installed
/vault-recall
- Read your context and instructions
- Compiled the vault recall
Read 01_Projects/shop-app/README.md and the last 3 chat logs (Oct 22 – Oct 26).
shop-app: Next.js 15 store with Prisma + Postgres, deployed on Vercel.
Decisions still in force
- Database sessions instead of JWT, so admins can revoke sessions
- Prices stored in cents as integers, never floats
Last session (Oct 26)
- Google login works; the prod redirect loop was
NEXTAUTH_URLon http
Open TODOs
- Add the GitHub provider
- E2E test for the sign-in flow
Heads-up: src/lib/money.ts now uses parseFloat, which goes against the "cents as integers" decision. Want me to fix it, or was that change intentional? Either way, the next vault sync will record it.
I'll follow these decisions for this session. What are we working on?
Connects securely to your tools. The creator never sees your data.
About this skill
Vault Recall reads your project's notes from your Obsidian vault when a session starts and gives you a 5–10 line recap:
- what the project is and what it's built with
- the decisions that still hold, and why they were made
- how the last session ended
- open TODOs and questions
After that, the agent follows those decisions instead of re-arguing them. If the code and the notes disagree, it trusts the code and points out the mismatch, so the next sync can record it.
Good to know
- Read-only: it never edits your vault.
- Works with any agent that reads SKILL.md.
- Made to pair with Vault Sync, which writes the notes.
Open source (MIT): https://github.com/ali-novruz/universal-vault-sync
- Needs notes to read. Run Vault Sync (or write the project README yourself) first. On a brand-new project it will say there's nothing yet.
- Reads the project README and the 3 most recent chat logs. It only searches older logs when you ask about something specific.
- The project is identified by the workspace folder name, so it must match the folder name in your vault.
- The agent needs permission to read your vault folder. Browser-only chats without file access can't use it.
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
Trust & safety
Security scan
Security scan passed 21 days ago
- Free with an Agensi account
Permissions
File Scopes
Read-only. Reads Markdown files in your vault's projects folder and the VAULT_PROJECTS_ROOT variable to find it. Never writes files, no network access, no shell commands.