minni-doctor

Diagnose Minni AI agent memory installation issues with evidence-based layer checks.

2|Updated Mar 31, 2026
One-click install
npx skills add https://github.com/infektyd/minni --skill minni-doctor
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: minni-doctor
Source: https://github.com/infektyd/minni/tree/main/plugins/minni/skills/minni-doctor
Command: npx skills add https://github.com/infektyd/minni --skill minni-doctor

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Minni is a local-first memory system for AI agents, but when it malfunctions, users often can't tell if the root cause is the daemon, vault, plugin wiring, hooks, or another component — leading to hours of guesswork and wasted troubleshooting time. This skill eliminates that frustration by running a systematic, layer-by-layer health diagnostic that identifies the exact root cause of any Minni issue, with concrete evidence to back up every finding, and never makes silent unapproved changes to your system.

Core Features & Use Cases

  • Ordered Layer-by-Layer Checks: Runs 8 sequential diagnostic checks (daemon/socket/DB, vault, plugin/MCP wiring, hooks, identity, inbox lifecycle, model providers, plans) ordered by dependency, so lower-layer failures don't cause false positives for higher layers.
  • Evidence-Based Reporting: Every finding includes concrete proof like file paths, RPC output, and file counts, so you never get a vague "looks fine" response without an actual check being run.
  • Clear Repair Routing: Automatically routes repair actions to the minni-install tool for config issues, or escalates engine-internal problems to a human, with no unapproved state changes.
  • Use Case: If your agent can't recall past corrections you gave it in a previous session, handoffs between Claude Code and Codex are failing, or new pages you write to your Minni vault aren't showing up in recall, this skill will tell you exactly which layer is broken and what to do about it.

Quick Start

Use the minni-doctor skill to run a full health diagnostic on your local Minni installation and get a prioritized list of all identified issues with evidence and fix recommendations.

Frequently Asked Questions about minni-doctor

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I diagnose why my AI agent memory daemon is failing to connect?

To diagnose AI agent memory daemon connection failures, run an ordered layer-by-layer health check that tests the daemon socket and database first. This sequential diagnostic approach prevents false positives in higher layers like vault or plugin wiring by identifying the root cause with concrete RPC output evidence.

What is the best way to troubleshoot vault misconfigurations in a local AI memory system?

The best way to troubleshoot vault misconfigurations is to run an evidence-based diagnostic that inspects file paths and counts within the vault layer. This automated check pinpoints the specific configuration error causing pages to fail appearing in agent recall.

Why does my agent forget past corrections after a session restart?

Agent memory loss after a restart can stem from plugin wiring errors, hook registration gaps, or identity layer delivery issues. A systematic diagnostic checks these dependencies in order, using file paths and RPC output to prove exactly which component failed to persist the session context.

Can I use this diagnostic tool to fix model provider misconfigurations automatically?

You cannot fix model provider misconfigurations automatically because the diagnostic performs no silent state changes. It routes repair actions to the appropriate minni-install workflow for configuration issues, or escalates engine-internal problems to a human operator.

How do I resolve inbox lifecycle deadlocks and plan injection failures in my AI agent?

To resolve inbox lifecycle deadlocks and plan injection failures, execute a full diagnostic sweep across all supported agent runtimes. It isolates the failure point using concrete evidence like file counts and RPC output, then routes the repair path without altering system state.