Skip to content
Back to skills

Monomind

ASecurity

Complete integration with Monomind — codebase knowledge graph (Monograph), persistent memory (L1/L2), Second Brain document vector search, autonomous agent orgs, and real-time statusline.

  • 21 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added September 10, 2026
ai-agentsbashnode

Works with

  • mcp

Security analysis

A100/100

Scanned September 28, 2026

npx -y skills add monoes/monomind --skill monomind --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Monomind?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Monomind
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/monoes-monomind/badge)](https://www.skillsdirectory.com/skills/monoes-monomind)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: monomind
description: Complete integration with Monomind — codebase knowledge graph (Monograph), persistent memory (L1/L2), Second Brain document vector search, autonomous agent orgs, and real-time statusline.
---

# Monomind Skill for Antigravity

Monomind extends Antigravity with a local codebase knowledge graph, persistent cross-session memory, semantic document search, autonomous agent orgs, and a real-time system statusline.

## Core Rules & Workflow Integration

### 1. Codebase Knowledge Graph (Monograph)
Before searching or modifying code touching 3+ files or exploring a new module:
- Call `mcp__monomind__monograph_suggest` to discover relevant files, bridge nodes, and isolated nodes.
- Call `mcp__monomind__monograph_query` for symbol/keyword search, with optional one-hop neighbor-expansion reranking.
- Call `mcp__monomind__monograph_impact` before editing to compute the exact blast radius of a symbol or class change.

### 2. Persistent Memory Loop
- **Recall**: Use `mcp__monomind__memory_pattern-search` to query past learnings, architectural rules, and session history before reinventing solutions.
- **Feedback**: After using memory search, call `mcp__monomind__memory_feedback` with entry IDs and outcome quality to continuously train search ranking.
- **Store**: Distill durable insights (entities, relations, lessons) into L2 memory via `mcp__monomind__memory_kg_ingest` or `mcp__monomind__memory_pattern-store`.

### 3. Second Brain (Document Vector Search)
- Project specs, handbooks, and notes indexed in `.monomind/data` or global brain (`~/.monomind/global-brain`) can be queried using `mcp__monomind__knowledge_search` or `npx monomind doc search -q "<query>"`.

### 4. Autonomous Orgs Daemon
- Run background role-based agent organizations using `monomind org run <name>`.
- Monitor active orgs via `monomind org status` or `monomind org list`.
- Check live agent questions using `monomind org questions <name>` and respond with `monomind org answer <name> <q-id> "<answer>"`.

### 5. Monomind Statusline
To inspect real-time project progress, domain coverage, graph node counts, and session cost metrics:
- Execute `node .gemini/helpers/statusline.cjs` (or `node .claude/helpers/statusline.cjs` / `npx monomind hooks statusline`).

```bash
# Example statusline execution
node .gemini/helpers/statusline.cjs
```

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.