Scored across 17 tools
There is apparent overlap among the evaluate family (evaluate_file/evaluate_project/evaluate_code/inspect_code) and the assess family (assess_worktree_change/assess_changeset/assess_improvement/assess_snapshot), but the descriptions explicitly delineate each by input type, baseline source, and intended loop, making selection reasonably clear. A few boundaries (inspect_code vs evaluate_file, compare_code vs compare_files) require careful reading but are documented.
Nearly all tools use a consistent topos_ + verb_noun pattern (evaluate_file, assess_changeset, begin_refactor, calculate_coverage, compare_code, generate_depgraph, get_doc, inspect_code). Minor deviations exist in noun-noun or bare-verb forms (depgraph_status, preference_walk, refactor), but the set remains predictable.
17 tools is on the heavier side but each maps to a distinct operation across evaluation, assessment, comparison, refactor advisory, dependency-graph management, and docs. The count is justified by the breadth of the quality/refactor domain rather than redundant surface.
The surface covers scoring (file/code/project), baseline assessment (git worktree, changeset, snapshot, proposed variant), comparison, refactor hotspot ranking, dependency-graph lifecycle, coverage, and documentation. Gaps are minor—no automated apply-fix or bulk multi-file evaluation beyond changeset, but core lifecycle is well covered.