Pipeline

The release-1 analytical pipeline is intentionally explicit and file-based.

Operations

• assess Input: one composition profile Output: one assessment_result

• coverage Input: one composition profile and one use-case file Output: one coverage_result

• compare Input: two assessment_result documents and optional alignment file Output: one comparison_result

• prioritize Input: one assessment_result, optional comparison_result, and optional list of coverage_result documents Output: one prioritization_result Consumes coverage gaps, maintainability findings, structural comparison observations, and alignment-gap observations when present.

• summarize Input: one or more result documents Output: one summary_result Produces a compact deterministic interpretation over existing results without rerunning analysis. The JSON result is canonical; the CLI may also render it as Markdown.

Result Documents

All analytical result documents use the same top-level envelope:

• result_type
• inputs
• content
• diagnostics

For comparison_result, the main analytical sections are:

• structural_comparison Metric deltas and ranked structural observations.

• alignment Present only when an alignment file is supplied. Includes aggregate counts, evaluated_pairs, and ranked_alignment_observations for unmatched declared equivalences.

For prioritization_result, the ranked target list may therefore include:

• coverage gaps
• maintainability findings
• structural comparison observations
• alignment gaps derived from declared but incompletely matched equivalences

For summary_result, the main sections are:

• headline
• result_types
• key_points
• follow_up_questions

Current release-1 analytical semantics are intentionally conservative, but they now include:

• direct contradiction detection for cardinality conflicts and datatype-versus-object-reference conflicts
• item coverage matching from SHACL property-path, owner-shape, and target-class evidence
• join coverage matching from shared owner shapes, explicit cross-shape object-reference paths, or record-level retrieval context where appropriate

Discovery

The tool exposes built-in artifacts through:

• schema
• vocabulary
• template
• capabilities

These discovery commands are intended for both humans and AI-agent skills.