architekt-jutra-code

Critiques requirements and interactively rebuilds them. Applies 4 checks — problem-vs-solution framing, observable behavior vs CRUD status (interactively reformulates into proper user stories), extensible signal map of hidden domain decisions, and rigid quantifier probing. Invoked ONLY on explicit request.

Interactive wizard for designing consistency units (aggregates). Guides the designer step-by-step through command extraction, pairwise conflict analysis, boundary decisions, and locking strategy. Invoke when the user asks about designing aggregates, consistency units, "projektowanie agregatów", "jednostki spójności", "jakie komendy się blokują", "granica agregatu", "współbieżna walka o zasoby", "rywalizacja o zasoby", "concurrent resource contention", or similar.

Reviews test code and suggests when testing strategy mismatches the problem class being solved. Detects output-based tests on integration code, interaction-based tests on pure transformations, missing state verification on stateful objects, and tests at wrong abstraction level. Invoke when user asks to review tests, "is my test strategy correct", "review my tests", "test strategy", "am I testing this right".

Answer questions about the AJ platform (host application, modules, entities, endpoints, plugins, extension points, manifests, storage strategies, host bridge, features) by querying the AJ knowledge graph in Neo4j via the `neo4j-aj-kb` MCP server. Use whenever the user asks "what / which / how many / list / show me" about platform structure, plugin wiring, REST endpoints, module contents, plugin manifests, extension points, filters, plugin data storage, or feature traceability — instead of grepping the codebase.

Verifies linguistic boundaries between bounded contexts by analyzing language.md files. Each language.md declares context role, relationships, and integration points — no separate context-map needed. Detects typical language leakage patterns (strings, events, API calls), proposes type-specific fixes (generalization, ACL, dependency inversion), and interactively validates with user. For single-module PRs, checks whether new concepts fit the module's linguistic space. Strictly read-only.

Lightweight research workflow that collects and cross-verifies information from multiple sources without synthesizing into a report. Produces raw findings with summary and cross-source verification. Reuses research-planner and information-gatherer subagents.

Critiques requirements and interactively rebuilds them. Applies 4 checks — problem-vs-solution framing, observable behavior vs CRUD status (interactively reformulates into proper user stories), extensible signal map of hidden domain decisions, and rigid quantifier probing. Invoked ONLY on explicit request.

Classify business requirements into one of 4 modeling problem classes (CRUD, Transformation & Presentation, Integration, Resource Contention). Runs a signal scan, asks targeted clarifying questions, and recommends an implementation approach with rationale. NOT an archetype — invoke when the user asks about modeling problem classes, "jaka klasa problemu", "jak to sklasyfikować modelarsko", "problem class", or similar. For archetypes (accounting, pricing), use the *-archetype-mapper skills instead.

Recognize and classify NLP metaprograms from utterances, written communication, or described behavior. Identifies which of 7 metaprograms are active, detects compound patterns, and suggests communication strategies adapted to the person's cognitive filters. Invoke when the user asks about metaprograms, communication style diagnosis, "jak rozmawiać z tą osobą", "jaki metaprogram", "jak się komunikować", or wants to analyze someone's communication patterns.

Distill bounded contexts by finding safe generalizations across domain concepts. Uses bidirectional linguistic analysis to detect where different things behave identically (generalization candidates) and where same-named things behave differently (context split candidates). Produces a context map with generalized and specific models.

Transform domain requirements into an accounting-style value flow model. Identifies resources, accounts, transactions, entries, reversals, validity periods, and allocation rules for any value-tracking system.

Scan domain requirements against all known archetypes in parallel. Launches one agent per archetype (accounting, pricing, party — extensible), collects fit/no-fit results, then merges into a single summary. Produces fit/ directory with individual archetype results and a combined report.

Transform domain requirements into a Pricing Archetype model. Identifies complexity level (1–9), designs Calculator layer, Component tree, Validity versioning, Applicability conditions, and context dimensions. Produces implementable model with explicit concept mapping and unmapped concepts sections.

Use when reviewing an AI agent's incident diagnosis output for precision, efficiency, and ownership awareness. Requires access to both the workspace artifacts and the ATIF trajectory.