Analysis patterns

Twelve patterns, each a method for a kind of architecture question: which templates to run, what gates them, and — the part that matters most — when to stop.

assets/patterns.json is the authoritative routing table. The prose under references/patterns/ is the depth behind each entry, and the suite tests the two against each other: every pattern has a reference file that exists, and every template it names is in the query catalogue. A pattern naming a template nobody ships is a routing instruction to nowhere.

python3 scripts/la-analyse plan --list-patterns
python3 scripts/la-analyse plan --question 'what depends on "Order Service"?' --mode impact-and-dependency

How a question reaches a pattern

flowchart TD
  Q["the question"] --> M{"--mode given?"}
  M -->|yes| P["that pattern, no scoring"]
  M -->|no| T["score every pattern:<br/>substring match for phrases,<br/>whole-word for single words"]
  T --> S{"best score > 0?"}
  S -->|yes| P
  S -->|no| Z["no pattern matched.<br/><small>orientation steps are still right;<br/>then name a pattern or say this<br/>package has no method for it</small>"]

A zero score everywhere is a real answer, not a failure: the question does not look like anything this skill has a method for.

The lists below are ordered candidate checklists, not mandatory execution scripts. Analyse de-duplicates queries, makes definitions conditional, records refused candidates as decisions, and preserves review barriers before dependent work. Stop when the evidence is sufficient; do not execute every candidate to complete a checklist. Query remains the execution owner.

capabilities here is advisory

The authoritative gate is the query catalogue's own requires, checked at render time. The pattern lists capabilities so a planner can say "this dataset cannot answer that" before running anything.

The twelve patterns

coverage-and-gaps

Coverage and gaps. Depth: references/patterns/coverage-and-gaps.md

Triggers: uncovered, coverage, incomplete, gap, gaps, missing, no owner, without an owner, unowned, how many have

Candidate templates, in evidence order:

• core/inventory
• core/elements-by-type
• core/element-detail
• core/coverage-gaps
• core/orphans

Stop when:

• the population is empty, which produces zero gaps and looks like perfect coverage
• the gap concentrates in one model, which is a partial export rather than an ownership problem

cross-notation

Cross-notation questions. Depth: references/patterns/cross-notation.md

Triggers: reconcile, same system, same thing, both tools, across tools, bpmn to archimate, backstage to c4, identity

Candidate templates, in evidence order:

• core/identity-audit — gated on identity_assertions
• core/label-collisions
• core/inventory-summary

Capability gates: identity_assertions (core/identity-audit only)

Stop when:

• identity-audit is refused: stop the identity-dependent join, not independent parts of a multi-part question
• only label candidates remain: report identity as unresolved and continue independent claims with their own evidence

governance-and-decisions

Governance and decisions. Depth: references/patterns/governance-and-decisions.md

Triggers: decision, principle, standard, policy, exception, constraint, approved, governance, compliant

Candidate templates, in evidence order:

• core/discover-predicates
• core/element-detail
• core/classified-by
• core/provenance

Stop when:

• no predicate in this dataset carries decisions or policies at all
• the only status found describes the conversion rather than the architecture

taxonomy-classification

Taxonomy classification. Depth: references/patterns/taxonomy-classification.md

Triggers: classified under, classified as, classified, classification branch, taxonomy branch, taxonomy category, category members

Candidate templates, in evidence order:

• core/resolve-concept — resolution step, not an extra pattern query
• core/classified-by
• core/provenance

Stop when:

• no matching SKOS concept or scheme is present in the semantic graph — do not invent an IRI
• more than one concept or scheme matches the term — ask which one was intended
• the branch has no asserted classifications — do not infer that no elements belong to it

The question's classification wording selects this pattern; the quoted SKOS label resolves a concept after selection. A label by itself never chooses a template.

notation-taxonomy

Notation type taxonomy. Depth: references/patterns/notation-taxonomy.md

Triggers: notation taxonomy, type taxonomy, grouped by domain, grouped by aspect, element by domain, element by aspect, type category

Candidate templates, in evidence order:

• core/resolve-vocabulary-concept — resolution step over attached vocabulary
• core/elements-by-category
• core/provenance

Stop when:

• the notation's matching vocabulary is not attached — zero categories do not mean zero elements
• multiple schemes or categories match — review the exact scheme IRI
• the category groups OWL types, not asserted instance classifications

viewpoint-selection

Viewpoint selection. Depth: references/patterns/viewpoint-selection.md

Triggers: which viewpoint, what viewpoint, choose a viewpoint, suitable viewpoint, viewpoint catalog, viewpoint library, capability map viewpoint

Candidate templates, in evidence order:

• core/viewpoint-catalog — resolution step over attached vocabulary
• core/viewpoint-detail
• core/views — actual views, not viewpoint prescriptions

Stop when:

• the matching viewpoint catalog is not attached
• stakeholder, purpose or allowed classes do not fit the question
• a viewpoint declaration or conformance assertion is mistaken for SHACL conformance

impact-and-dependency

Impact and dependency. Depth: references/patterns/impact-and-dependency.md

Triggers: retire, replace, migrate, decommission, affected by, depends on, dependency, blast radius, impact

Candidate templates, in evidence order:

• core/resolve-element
• core/discover-relationship-types
• core/neighbours-qualified
• core/dependents-qualified
• core/reified-predicates — gated on rdf_reifies; names the predicates core/dependents-direct follows
• core/dependents-direct — gated on direct_rel_triples
• core/provenance

Capability gates: direct_rel_triples (core/dependents-direct only), rdf_reifies (core/reified-predicates only)

Stop when:

• the relationship semantics on a path stop supporting the claim
• reachability has been established but criticality has not - they are different questions
• the next hop needs data the models do not carry, such as traffic or failure history

lifecycle-and-portfolio

Lifecycle and portfolio comparison. Depth: references/patterns/lifecycle-and-portfolio.md

Triggers: duplicate, duplicates, overlap, rationalise, rationalize, lifecycle, deprecated, portfolio, strategic importance

Candidate templates, in evidence order:

• core/lifecycle — gated on element_lifecycle
• core/define-term
• core/label-collisions
• core/identity-audit — gated on identity_assertions
• notation/leanix/factsheets

Capability gates: element_lifecycle (core/lifecycle only); identity_assertions (core/identity-audit only)

Stop when:

• equivalence rests on labels alone - say the question is not answerable yet
• lifecycle is absent for most of the population, so the listed subset is not a portfolio view

model-contents

What a model contains. Depth: references/patterns/model-contents.md

Triggers: what is in, what does this model cover, take part, takes part, participate, who is involved, contents, components

Candidate templates, in evidence order:

• core/inventory-summary
• core/resolve-model
• core/define-term
• notation/bpmn/process-components
• notation/bpmn/process-flow
• notation/c4/containers

Stop when:

• the model is drawn at a level of detail that hides what was asked about
• a participant has been listed - that is a drawn intent, not a running system

model-quality

Model quality. Depth: references/patterns/model-quality.md

Triggers: can we trust, trust, quality, complete, wrong with, orphan, orphans, conformance, conform

Candidate templates, in evidence order:

• core/inventory-summary
• core/orphans
• core/coverage-gaps
• core/provenance
• core/identity-audit
• core/validation-summary — gated on validation_in_graph

Capability gates: validation_in_graph (core/validation-summary only)

Stop when:

• the question turns out to be about conformance against shapes - hand it to linked-archi-validate
• a verdict is available but its target-class coverage is not, which makes the verdict unreadable

traceability

Traceability. Depth: references/patterns/traceability.md

Triggers: which capability, supports, realises, realizes, implements, serves, cross-layer, end to end, traceability, trace

Candidate templates, in evidence order:

• core/resolve-element
• core/traceability
• core/coverage-gaps

Stop when:

• no cross-layer relationship type exists in this dataset, which is a finding
• a missing link has been identified - report it separately from a negative assertion

views-and-documentation

Views and documentation. Depth: references/patterns/views-and-documentation.md

Triggers: diagram, diagrams, drawn, documented, view, views, what is new in

Candidate templates, in evidence order:

• core/views — gated on views_graph
• core/view-contents — gated on views_graph
• core/view-usage — gated on views_graph
• core/view-diff — gated on views_graph

Capability gates: views_graph (every template in this pattern)

Stop when:

• the notation carries no diagrams at all, which is always true of Backstage and LeanIX
• a difference between diagrams has been found - it is not yet a difference in the architecture