cc-tree: a radial-tree plugin for auditable Claude Code exploration

cc-tree is a Claude Code plugin that turns a single prompt — brainstorm, critique, compare designs, or audit code — into a branching exploration tree where each node is expected to cite evidence and the run continues until branches stop producing new high-value directions. It is aimed at developers and technical leads who find ordinary chat output hard to review or act on. The README frames the target problem as five recurring failures: shallow coverage, deferred leaves, pseudo-divergence, convenient convergence, and unverifiable claims. The plugin’s stated inversion is fixed-breadth coverage, no deferrals, merged duplicates, a checkable convergence test, and a file:line or URL behind every claim. One engine, four presets The core idea is a tree grown outward from one root, which is your input. Every expansion tries 12 framings, each child is derived and scored, and only children marked as advancing are expanded again. That means branches can stop at different depths, and the run ends on convergence rather than on a preset node budget. The README illustrates this with a single run of 49 nodes and 45 terminal tips, drawn by a generator tool that, according to the project, refuses to draw a tree the engine could not produce. Four presets reuse the same loop with different vocabulary and deliverables, plus a chaining command: /cc-tree:brainstorm for divergent exploration, producing shortlist.md /cc-tree:attack for adversarial critique, classifying leaves as CONFIRMED, MARGINAL, or REFUTED and producing confirmed.md /cc-tree:design for design-space exploration across options, trade-offs, reversibility, and cost, producing options.md /cc-tree:code-audit for threat-model, contract, and cross-file issues, producing findings.md with file:line and