This directory contains implementation and operational reference documentation for the Minsky development workflow tool. Strategic memos, position papers, and incident records live in Notion (Minsky workspace); per-task history lives in the Minsky task DB and GitHub PRs.
- Main README — project overview, installation, and basic usage
- CONTRIBUTING.md — TypeScript dev setup, testing, and the project record stack
- CLAUDE.md — instructions for AI agents working in this codebase
- Bun Optimization Setup — disk space optimization (60-90% savings)
- Domain Concepts — Core Minsky concepts (Repository, Session, Workspace)
- Theory of Operation — How cybernetic theory maps to code modules (VSM, environmental pre-delegation)
- Architecture Overview — System architecture, command registry, persistence, DI
- SessionDB Migration Guide — Migrate between SQLite and PostgreSQL backends
- SessionDB Troubleshooting Guide — Solutions for common SessionDB issues across all backends
- Postgres Persistence Configuration — Pool sizing, retry policy, and graceful shutdown for the Postgres backend
- Multi-Backend User Guide — Complete guide to the multi-backend task system
- Multi-Backend Quick Reference — Command reference and cheat sheet
- Tasks Dependencies — Task dependency model and CLI
- Shared Command Registry — How CLI and MCP interfaces share the same command definitions
- MCP Server — How to start and inspect the MCP server locally
- MCP Migration Guide — Notes on MCP method/parameter migrations
- MCP Schema Composition Guide — Schema patterns for MCP tools
- MCP Signaling Spike Findings — Findings from the mt#1315 signaling spike
- PR Workflow — Pull request preparation and approval workflows
- Configuration Guide — Configuration overview
- Configuration Validation Guide — Validation patterns
- Repository Configuration — Repository backend configuration
- GitHub App Bot Setup — Configure GitHub App bot identity (
minsky-ai[bot]) for automated PR reviews - GitHub Issues Backend Guide — GitHub Issues task backend setup
- Development Workflow — Pre-commit hooks, testing, and quality gates
For all testing decisions (where to put tests, how to mock, what to test) see the
testing-guide skill (.claude/skills/testing-guide/SKILL.md). Mocking utilities
live in src/utils/test-utils/. Architecture-level docs:
- Test Architecture Documentation — Architecture of the test infrastructure
docs/architecture/— ADRs and architecture deep-dives.
- Memory Migration — Memory subsystem migration notes
- Domain Schema Architecture — Domain schema design
- ASCII-Only Code Symbols — see CLAUDE.md (project-wide ASCII rule)
- Variable Naming Guide — Variable naming conventions
as unknownPrevention — Guidelines against unsafeas unknowncasts- Stale Reference Checklist — Checklist for catching stale references during edits
- Logging — Logging configuration and patterns
- Deploy Minsky to Railway — Railway deploy procedure for hosted Minsky MCP
- Session Management:
minsky session start|list|get|delete|dir|update|pr|approve - Task Management:
minsky tasks list|get|create|status|spec|migrate - Git Operations:
minsky git clone|branch|summary|prepare-pr - Rules Management:
minsky rules list|get|create|update|search|delete - Configuration:
minsky config list|show - Project Setup:
minsky init - MCP Server:
minsky mcp start
- Task Backends:
minsky,github-issues - Repository Backends:
local,remote,github - Session Storage Backends:
sqlite(local database),postgres(server database)
- Main Configuration:
.minsky/config.yaml - Task Specs: stored in the Minsky task DB (browse via
mcp__minsky__tasks_*) - Rules Directory:
.cursor/rules/(Cursor format) or.minsky/rules/(generic) - Session Storage:
~/.local/state/minsky/sessions/ - AI Agent Memory:
~/.claude/projects/.../memory/(per-machine, not committed)
When updating documentation:
- Keep examples runnable against the current codebase. If a doc claims a tool/utility exists, verify it does.
- Update cross-references when moving content; broken links are surfaced by the stale-reference checklist.
- Prefer linking from this index over duplicating content in multiple places.
- Per-machine, ephemeral, or task-specific documentation does not belong here — that lives in tasks (DB), PRs (GitHub), or memories.
For per-cluster cleanup history, see PRs filed under mt#1319 / mt#1360.