Mirror Extensions

A monorepo of stateful extensions for Mirror Mind.

Mirror Mind is a configuration and memory framework for a Jungian mirror AI. It ships with two extension kinds: prompt-skill (markdown-only workflows) and command-skill (extensions with their own SQLite tables, CLI subcommands, and Mirror Mode integration). This repository hosts a collection of command-skill extensions that the author uses day-to-day and is happy to share.

Each subfolder is a self-contained extension. They install one at a time through the framework's standard CLI:

git clone https://github.com/mirror-mind-ai/extensions ~/Code/mirror-extensions

python -m memory extensions install <extension-id> \
  --extensions-root ~/Code/mirror-extensions

The target mirror home is resolved from MIRROR_HOME / MIRROR_USER in the active environment. Pass --mirror-home <path> explicitly only when overriding the default.

Extensions in this repo

Folder	Status	What it does
finances/	✅ Complete (11/11 stories)	Personal and business finance tracking: accounts, transactions, balance snapshots, recurring bills, runway, monthly cash flow, and a financial_summary capability that injects live numbers into Mirror Mode for a finance-aware persona. Complete CLI surface: register accounts, record balance snapshots, manage recurring bills and categories, import bank statements (OFX) and credit card statements (CSV), list and filter transactions, compute burn and runway under different assumptions, and migrate from a legacy mirror SQLite database. See finances/docs/user-stories/ for the full story trail.
testimonials/	✅ Complete (5/5 stories)	Customer testimonials with LLM-assisted structuring and semantic search. add accepts free text and uses the framework's LLM router to extract author, source, product, quotable highlight, and tags; list filters by product / author / source; search ranks by cosine similarity over embeddings; migrate-legacy imports a legacy archive with embeddings preserved verbatim; recent_testimonials injects semantically relevant testimonials into Mirror Mode for bound personas. See testimonials/docs/user-stories/.
maestro/	🚧 First operational slice	Mirror extension that operates the Ariad method (integral agentic development). Commands: doctor for Builder Mode readiness, init for new projects, adopt for existing ones (with safe write and --dry-run), update for report-only comparison against canonical templates. Ariad is the method (lives in its own canonical repo); Maestro is how Mirror runs it.

Requirements

• Mirror Mind with the stateful extension system (CV14.E1, command-skill kind), available on main since 2026-05-11.
• Python 3.10+.
• uv for running the framework's commands.

Repository layout

.
├── LICENSE
├── README.md
├── finances/
│   ├── README.md            -- per-extension entry point
│   ├── skill.yaml
│   ├── SKILL.md
│   ├── extension.py
│   ├── migrations/
│   ├── src/
│   ├── tests/
│   └── docs/
└── testimonials/            -- (placeholder, not started)

Every extension follows the layout recommended by the framework's Authoring Guide.

Running tests

Tests for each extension live under that extension's tests/ directory. They depend on Mirror Mind being importable as memory, which today means running them from inside a Mirror Mind checkout's uv environment:

# From the Mirror Mind repository root:
uv run pytest ~/Code/mirror-extensions/<extension>/tests/

Once Mirror Mind distributes a published package, the tests will run from the extension's own pyproject without that hop.

Contributing

These extensions started as a personal port of features from an earlier mirror prototype to the current framework. Issues and PRs are welcome, especially:

• new extensions that follow the framework's command-skill contract and pass a clean review against the Authoring Guide;
• fixes to the existing extensions;
• documentation improvements.

Open an issue first when in doubt about scope.

License

MIT © Alisson Vale.