Concepts
This page explains the mental model AlienMonster is built on: what strategy data is, how it is organized, and how it moves between the server, your repo, and your coding agent.
The strategy repository
Strategic context — decisions, assumptions, risks, principles — is data you can version, query, and feed to agents. Treat it that way and a lot of decisions get easier: agents stop re-litigating choices you have already made, and you do not lose track of why you made them.
That is the whole thesis.
Primitives
A primitive is one strategic fact. Examples:
- "We chose Paddle over Stripe because of EU VAT handling."
- "Assumption: users tolerate up to 500ms auth latency."
- "Risk: our GPU provider has no SLA for inference outages."
Each primitive has a dimension, a subtype, an ID, and a content block. The ID format is <DIM>-<SUBTYPE>-<NUM>— for example, DEC-DEC-001 for a decision primitive or RSK-ASMP-001 for a risk-dimension assumption subtype.
The structural detail is in Primitives.
Dimensions and clusters
AlienMonster organizes primitives into 12 dimensions, grouped into 4 clusters:
- Core — who, what, why (customers, product, vision)
- Operations — how, capability, money (delivery, team, revenue)
- Environment — market, context, risk (competition, regulation, threats)
- Evolution — evidence, direction, decisions (signals, roadmap, choices made)
Each dimension has subtypes that narrow the shape. For example, the decisions dimension has subtypes for decision, value_principle, and pivot.
The full taxonomy — all 12 dimensions, all 93 subtypes, and the ID prefix table — lives in src/lib/taxonomy.ts.
Powers — curated context for agents
Without strategic context, coding agents guess. With too much, they drown in it. Powers are the middle path: focused markdown files that aggregate your primitives by strategic area, so agents can load only what's relevant to the task in front of them.
AlienMonster ships five powers — target-user, business-model, market-position, tech-decisions, risk-register. Each renders automatically from the primitives that feed it. You don't create powers directly; if you answer the Facilitator's questions about users, the target-user power renders on its own.
Agents activate powers on-demand via the get_relevant_powers MCP tool, which matches your task description and file paths against each power's inclusion rules. Read more in Powers.
Sessions, primitives, artifacts, and powers
Four kinds of data. Don't confuse them.
- Sessions are the conversations. A session is a bounded exchange with the Facilitator — a Strategy Interview, a Brainstorm, or a Quick Update.
- Primitives are the structured facts produced by sessions. A session might create 5 or 15 primitives.
- Artifacts are longer-form outputs — a spec, a pitch, a design doc — that coding agents or the Facilitator produce from primitives.
- Powers are curated markdown files that aggregate primitives into focused context for coding agents. They render from your primitives; you don't author them.
Sessions are covered in Session types. Artifacts surface in the workspace view. Powers live at Powers.
Why markdown and JSON
Every primitive renders two ways.
The markdown version is what people read — for example, .alienmonster/decisions.md lists every decision-dimension primitive as a readable heading plus prose.
The JSON version (.alienmonster/strategy.json) is what agents consume when they want structured data rather than prose.
Same content, two audiences. Humans get legible text; agents get a parseable structure.
The .alienmonster/ mirror
Your strategy lives authoritatively on the AlienMonster server. The .alienmonster/ directory on disk is a cache. It is fast to read for agents that want files, but it is not the source of truth. Sync commands keep the cache aligned with the server; export can always rewrite it from scratch.
Read more in The .alienmonster/ directory and Sync model.