Skip to content

lockstride/dmatrix-extensions

Repository files navigation

dMatrix Extensions

Bring dMatrix decision-capture to your favorite AI assistant. dMatrix notices when you're weighing alternatives, evaluating trade-offs, or making a decision — and offers to capture it as a structured, auditable decision matrix.

Pick your assistant

Assistant Guide One-step setup?
Claude (Code, claude.ai, Desktop, Cowork) claude/README.md ✅ One install in Claude Code — the app adds one click (plugin → Connectors → Install)
OpenAI (ChatGPT & Codex) openai/README.md — two quick steps
Google (Antigravity CLI) google/README.md ✅ Antigravity plugin

Every setup has two parts — you need both

  1. The behavior — instructions that teach the assistant how to capture a decision (when to offer, how to draw out options, criteria, and scores).
  2. The connection to dMatrix — linking the assistant to your dMatrix account so it can actually create and save your decision matrix.

Both parts are required. With only the behavior, the assistant can talk about your decision but can't create or save anything in dMatrix — so it ends up offering help it can't deliver. Always set up both. Some assistants do it in one step; others take two. Each guide walks you through it. You'll need a free dMatrix account — sign up at dmatrix.com.

What you get

Once set up, the assistant will:

  • Detect decision signals — comparative evaluation, trade-off language, criteria-based reasoning, explicit "help me decide" framing.
  • Offer to capture — a brief, non-blocking suggestion, never interrupting your flow.
  • Build a structured matrix — options, criteria, scores, and weights, extracted from what you actually said.
  • Track provenance — every value links back to its source, with confidence levels and citations.

Detection stays conservative — dMatrix only offers to capture on a clear decision signal, and never interrupts.

You can always ask

You don't have to wait for the assistant to notice a decision. At any point, type:

/dmatrix:decide

This works in Claude Code, Codex CLI, and Antigravity CLI. It tells the assistant to capture a decision right now — even if you're mid-conversation, or looking back at a decision you already made. The assistant will review what you've discussed and build a matrix from it.

Use it when:

  • You realize a past conversation contained a decision worth capturing.
  • You want to structure a decision you're actively working through.
  • The assistant didn't offer on its own and you'd like it to.

Repository structure

shared/canonical-prompt/source.md   # The single source of truth (canonical behavior)
.claude-plugin/ · .mcp.json         # Claude plugin manifest, self-vending marketplace + MCP wiring
                                    # (this repo is its own single-plugin marketplace: source "./")
skills/decide/SKILL.md              # The dMatrix Skill (Claude / standalone)
claude/README.md                    # Claude install guide (all surfaces)
openai/                             # ChatGPT paste-in + Codex (AGENTS.md, config.toml)
google/antigravity/dmatrix/         # Antigravity CLI plugin bundle
mcp/connectors/                     # Host-agnostic MCP connector reference

How it works

Every behavior-bearing artifact is generated from one source, shared/canonical-prompt/source.md, by the renderer pipeline in the dMatrix monorepo (packages/canonical-prompt). The pipeline injects the canonical body verbatim into each surface — only the per-provider wrapper differs — and stamps a content hash (version_hash) so every artifact stays provably in sync. To re-sync after the canonical prompt changes, the pipeline regenerates this whole repo in one run; the artifacts here are not edited by hand.

Support

About

Connect your AI assistant to dMatrix. Detects when you're weighing a decision and captures it as a structured, auditable matrix.

Topics

Resources

License

Stars

0 stars

Watchers

0 watching

Forks

Packages

 
 
 

Contributors