Claude Skills by provenmap

How to adopt an architecture decision (ADR) into ProvenMap and measure the estate against it — the /adopt-adr arc. Use when the org has made a decision, standard, or policy that boards should be governed by and checked against. Key capabilities: ADR normalization, blast-radius sweep per aspect family, the decision grill, the durable decision board, compliance insight batches, federated per-app remediation work items.

Take a new app from placed to build-ready — the readiness bar, the spec grill that lands sequenced work items, and skill selection. Use when a convert_node_to_app ran with observationType 'new_app', when /prepare-app runs, or when the architect asks to "prepare/spec the app", "get it ready to build", "choose its skills". Key capabilities: the three-point readiness bar (founding work item, spec set, skills), spec decomposition with dependsOnSlugs sequencing, judgment-based skill proposal over ...

The architect workbench identity — how to work over the ProvenMap MCP server as the board orchestrator brain. Use in every architect session, before any board, work item, or insight work. Key capabilities: role and capabilities, the token scope model and write fence, board taxonomy and routing rules, the workflow routing table for open-ended asks, the working copy (journal → preview → commit), passive review (governance is born at commit), formatting norms, canonical error vocabulary.

Bootstrap an existing-but-empty board below root into an implementable app design. Use when /board meets an empty app board (bound governing/reference, pre-first-push, or unbound with an app-archetype owner node) or an empty plain layer, or when /start routes a "shape/prepare/initialize this empty board" ask. Key capabilities: the two entry styles (architecture-first, pages-first via author_pages), intended-aspect authoring (author_endpoints, author_tables, author_channels, author_boundary_ru...

How to orient on and evolve a ProvenMap board over MCP — read the layered graph, analyze it, and make governed diagram edits. Use when exploring a board, answering architecture questions, assessing quality, or adding/updating/removing nodes and edges. Key capabilities: orientation sequence, board-class detection (landscape/app/layer), question-scoped reading and the answer-mode heuristic, cross-board navigation (layers), analysis types and assessment criteria, the diagram tool contract (slugs...

How to style a ProvenMap board so it reads professionally — composition, size hierarchy and icons first, semantic tokens sparingly. Use when authoring or beautifying any board (inline in setup-workspace/new-app/board writes, or via /style-board). Covers the signals→plan→validate→apply pipeline and the three apply_* tools.

How /discover turns a scored graph reading into insights and context boards worth showing — the plan and its menu, the recommended set, the briefs, the authoring agents, the push order and the report. Use when running /discover, when authoring one discover brief, or when reading a discover plan or ledger. Key capabilities — the two answer shapes, the wow levers, the network-free agent contract, the push-in-order rule, re-runs that replace.

How to write the words of a ProvenMap insight or context board — the one contract for every flow that authors them. Use whenever you fill an InsightDraft's name, insight, impact, advice or trail notes, a push's info line, or a context board's name, description and node notes (/insights, /discover, an analysis recorded from the architect workbench). Covers who the reader is, the five writing rules with a bad/good pair, and the text-field guide with limits.

How to review ProvenMap insight batches and turn insights into work, how to run a structured assessment (/assess), and how to record a session analysis as a draft batch. Use when listing or reading insights, deciding what to promote to work items, running a periodic review, or capturing an architecture review performed in-session. Key capabilities: the batch→insight drill-down, insight ids, promotion to draft work items, the assess arc (frame→sweep→review→record), template honesty, create_ins...

How to model an org's digital estate on the ProvenMap root landscape — the inception doctrine, archetype selection, and the binding handoff. Use when bootstrapping an empty workspace (/setup-workspace), placing a new system (/new-app), or editing the root landscape in a /board session. Key capabilities: the modelling doctrine (model what exists, one node per system), the source gate and shallow repo scan (--scan-repos), the Aspire AppHost landscape proposal, per-system kinds, app archetypes a...

How every ProvenMap command closes — the Outcome you write from the script's brief. Use at the last step of any command or routed workflow, when the step says "Done · Left · Next" or names --brief. Covers the fixed shape, the eight writing rules with examples, how to read a brief (state, gates, candidates, hand-offs), the one-line form, and what never appears.

How to author a ProvenMap playbook — an org-level guided track for one job, composed from steps the platform already watches. Use when the architect wants to turn a job ("replace the stale doc", "run our ADR process", "onboard a new team to the map") into a playbook others in the org can start, or wants one like an existing playbook with changes. Key capabilities: the authoring interview (job → overlap/template → shape → human-only parameters → read-back → draft), composition rules (building ...

How to RUN a ProvenMap playbook from the terminal — pull the playbook's compiled skill file, work its steps in order through the platform's MCP tools, and let the server prove each one. Use when the architect wants a playbook done rather than authored ("run the migration playbook", "walk me through the ADR process", "do the inherited-system one on this board"), or asks what a live run needs next. Key capabilities: the run loop (list → start → pull the skill → step → re-read), which steps an a...

How to draft a ProvenMap bug report that describes the problem without naming the user's project. Use at /report-bug's drafting step. Covers what to keep (the scenario), what never goes in (the user's names and content), the report template, and how to generalise a failure.

How to author, anchor, and manage ProvenMap work items — the living specs that connect architecture decisions to delivery. Use when creating work items, authoring requirements against a board, managing the work item queue (transition, assign, delete), or turning insights and board changes into work. Key capabilities: the work item lifecycle, DRAFT-only stance, the impact→attach→describe authoring loop, the authoring interview, anchor recording with the platform's verbs, one-app rule (referenc...

Extracts a repository's API surface (endpoints, protocols, auth, params) into an ApiSurfacePayload for /adopt --api. Use when adopting the api.surface aspect. Reads OpenAPI specs, NestJS/Express controllers, tRPC routers, or GraphQL SDL — never calls the endpoints.

Heuristics for identifying when a codebase needs a NEW archetype on the server, or when an EXISTING archetype is being stretched/misapplied/duplicated. Used by /analyze-archetypes, and by /analyze when recording the archetype gaps it hit. Defines what makes a good proposal, what to avoid, and how to phrase rationales.

The archetype field contract — which node/edge properties a codebase may fill, which belong to the architect, and how a value reaches the board. Use when authoring board nodes with `attributes`, when a sync reports dropped attribute values, or when deciding whether a property can be evidenced from code at all. Key capabilities: the three ownership tiers, the evidence bar for a model-filled field, the closed-vocabulary rule.

Extracts a repository's authorization model (roles, resource types, permissions, policies) into an AuthzRegistryPayload for /adopt --aspect authz.registry. Use when adopting the authz.registry aspect. Reads CASL rule definitions, Cerbos policy YAML, OPA Rego policies, Oso Polar files, AWS IAM policy documents, or DB-backed RBAC table schemas — never evaluates a policy or queries live rows.

Author a styling plan for an analyzed board — deterministic signals, model judgment, offline validation. Use when /analyze finishes a board, when running /restyle, or when the user asks to beautify a board or diagram. Covers composition, size presets and icons, and the sparing use of semantic tokens.

Bundled Node.js CLI scripts for syncing codebase analysis to the ProvenMap Claude Code Plugin API. Provides pmap-sync.js for smart diff-based sync, pmap-archetypes.js for server archetype caching, pmap-boards.js for board management, and pmap-adopt.js for aspect adoption. Self-contained — no npm install required.

Analyzes codebase architecture across multiple languages to extract components, relationships, and tech stacks for C4-style visualization. Use when user asks to "analyze codebase structure", "detect tech stack", "find components", "identify frameworks", "map architecture", or "analyze monorepo". Supports JavaScript/TypeScript, Python, Java, Go, C#/.NET, Ruby, and Rust.

Extracts a repository's database schema (tables, columns, foreign keys, indexes) into a DatabaseSchemaPayload for /adopt --db. Use when adopting the database.schema aspect. Reads Drizzle, Prisma, TypeORM, SQLAlchemy, raw SQL migrations, or a committed pg_dump — never a live database.

Extracts a repository's event/messaging catalog (channels, event types, subscriptions) into an EventCatalogPayload for /adopt --aspect event.catalog. Use when adopting the event.catalog aspect. Reads AsyncAPI specs, Kafka/kafkajs configs, AWS SQS/SNS/EventBridge infra-as-code, RabbitMQ, Google Pub/Sub, Redis Streams, or NATS declarations, and .NET event buses (IEventBus-style RabbitMQ, MassTransit, Dapr pub/sub) with their integration-event records — never connects to a live broker.

How /ground decides which board nodes are substantiated by which repo documents — the citation quality bar, drift handling, and per-genre reading priority. Use during /ground's evidence-proposal step, or whenever judging if a document backs a node's claim. Defines what counts as substantiation, anchor/excerpt discipline, and how to treat drifted or unlinked evidence.

Bundled Node.js CLI scripts for fetching server-defined insight skills and pushing analysis results to the ProvenMap Claude Code Plugin API. Provides pmap-insights.js for skill caching, insight persistence, and server push. Self-contained — no npm install required.

How /monitor turns raw monitoring output (Sentry issues, CloudWatch alarms, cost anomalies) into normalized signals, correlates them with board elements, and shapes work-item-ready insights. Use when running /monitor, normalizing vendor monitoring data, or deciding how an operational signal maps onto architecture nodes/edges. Defines the signals schema, stable-id rules, and the teach-once mapping loop.

Extracts a repository's frontend routes/pages (route tree, render tree, data dependencies, guards, forms) into a UiPagesPayload for /adopt --aspect ui.pages. Use when adopting the ui.pages aspect. Reads Next.js App/Pages Router, Remix, TanStack Router, React Router, SvelteKit, or Nuxt route files — never renders or starts the app.

How to build an app from the ProvenMap platform's spec — reading compiled skills as the primary spec, turning designed board elements into components, materializing aspect snapshots (DB schema, API surface) as migrations and contracts, and closing the loop so the board maps the built app. Carries the planning and implementation steps /build delegates here. Used by /build.

Integrates with the Claude Code Plugin API to push architecture data for visualization on ProvenMap workboards. Use when user asks to "sync to ProvenMap", "push architecture to portal", "configure API", "connect to portal", "upload nodes", or "sync edges". Provides API patterns, authentication, and sync protocols.

Interpret pmap-inspect sessions — screenshots, picked components, annotations — and turn them into precise code changes or work item language. Use when reading .provenmap/inspections/ manifests, after /inspect completes, or when component identity in a session needs improving. Covers the identity ladder, ui.page correlation, and adopting compile-time source attributes.

Review an architect-authored work item against current code reality before implementing it, then classify the gap as clean, minor, or blocking. Used by /work-items Step 4.5 to decide whether to implement, or bounce the work item back to the architect for clarification. Provides the gap-table method, the blocking criteria, and how to draft the bounce-back note.

Extracts a repository's outbound service calls (typed HTTP clients, gRPC clients, service-discovery addresses) into an ApiClientsPayload for /adopt --aspect api.clients. Use when adopting the api.clients aspect. Reads .NET AddHttpClient/AddGrpcClient registrations, Aspire WithReference, Refit, fetch/axios base URLs, Feign, gRPC stubs — never calls the services.