feat(review): role-driven UX cutover + TUI (PR 3 of 3)

[peyton-alt]

Stacked on #1370 (PR 2: codex review correctness), which is stacked on #1241 (PR 1: role + setup foundation). Diffed against review-codex.

The command-layer redesign of entire review. Originally bundled with codex correctness; that agent-layer work was peeled into #1370 so this PR is purely the cobra/huh/TUI surface (no agent/* churn here).

What this adds

• Role-driven dispatch: drops the legacy in-flow picker; reviewer membership is decided up front by saved roles (entire review setup) or one-off --reviewers/--fixer flags. Invoker-aware non-interactive fallback; recursion guard on ENTIRE_REVIEW_SESSION. Deletes the spawn-time multi-agent picker.
• entire review fix promoted to a real subcommand (legacy --fix hidden). FixerOf(s) is the single source of truth for the fix agent.
• Pre-launch staging view: scope banner + itemised checkpoints/sessions + optional per-run prompt in one styled huh form before fan-out.
• Inline post-review fix prompt [Y]es · [s]elect · [n]o · [A]lways; unified navigable source⇄findings fix picker (aggregate selectable).
• Live-token display (TUI input-only during streaming) and muted markdown palette for dense review/synthesis output.
• Opt-in roles (new agents default Skip) with a ≥1-reviewer guard.
• Review-feedback fixes: codex skill seeding via seedDefaultSkills, codex-only completion footer, inline_prompt doc/test naming. Removed a stray .antigravitycli config committed in error.

Verified: build + vet clean, mise run lint 0 issues, full unit suite green (under TZ=UTC; the one local failure is a pre-existing timezone-sensitive auth test, green on CI).

🤖 Generated with Claude Code

[cursor]

Cursor Bugbot has reviewed your changes and found 4 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit ba16605. Configure here.}

[Copilot]

Pull request overview

This PR completes the entire review redesign by improving codex review correctness (skill invocation/discovery, live token tracking, manifest inclusion) and upgrading the review/fix UX (role-driven setup, staging view, unified fix picker, muted rendering), while keeping review configuration clone-local for privacy.

Changes:

• Add per-agent review knobs (model, reasoning_effort) and thread them through reviewer run config (notably codex -m / -c model_reasoning_effort=).
• Replace spawn-time picker flows with role-driven configuration (entire review setup), pre-launch staging/inline prompts, post-review fix prompt, and a unified interactive fix picker.
• Improve codex integration: discover $skill-form skills, tail rollout JSONL for live token counts, and ensure codex output is included in fix manifests even without hook-tagged sessions.

Reviewed changes

Copilot reviewed 47 out of 47 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
cmd/entire/cli/settings/settings.go	Add model / reasoning_effort to persisted ReviewConfig and update zero-check logic.
cmd/entire/cli/review/types/reviewer.go	Thread per-spawn model/effort overrides into reviewer RunConfig.
cmd/entire/cli/review/tui_model.go	Calm dashboard header styling; improve token column rendering for input-only vs final totals.
cmd/entire/cli/review/tui_model_test.go	Add coverage for token display branches in the dashboard.
cmd/entire/cli/review/synthesis_sink.go	Switch synthesis rendering to muted markdown palette.
cmd/entire/cli/review/setup.go	Implement role-first review setup, invoker detection, opt-in reviewer defaults, banner improvements.
cmd/entire/cli/review/setup_test.go	Update setup tests for clone-local persistence + opt-in role defaults + banner annotations.
cmd/entire/cli/review/setup_invoker_test.go	Add tests for invoker env-sentinel detection.
cmd/entire/cli/review/post_review.go	Add staging view + inline per-run prompt + post-review fix prompt/launcher + clone-local persistence for “Always”.
cmd/entire/cli/review/post_review_test.go	Add tests covering findings counting and post-review fix prompt behavior.
cmd/entire/cli/review/picker.go	Remove legacy picker/setup flows; codex skill verification becomes advisory; keep selection helpers.
cmd/entire/cli/review/picker_test.go	Pin codex advisory skill verification behavior.
cmd/entire/cli/review/multipicker.go	Remove spawn-time multi-agent picker (roles drive reviewer membership).
cmd/entire/cli/review/multipicker_test.go	Remove tests for deleted multi-picker.
cmd/entire/cli/review/manifest.go	Include session-less sources with output (codex exec) in fix manifests; improve filename handle selection.
cmd/entire/cli/review/manifest_test.go	Update command strings and add coverage for “output but no session” manifest inclusion.
cmd/entire/cli/review/inline_prompt.go	Add single-key reader helper for post-review fix prompt.
cmd/entire/cli/review/inline_prompt_test.go	Add tests for ReadSingleKey behavior.
cmd/entire/cli/review/flags.go	Add --reviewers / --fixer one-off override resolution + merging with saved skills/prompt.
cmd/entire/cli/review/flags_test.go	Add tests for flag override resolution + merge behavior.
cmd/entire/cli/review/fix.go	Add entire review fix subcommand, recursion guard, unified interactive fix picker, severity parsing improvements, and FixerOf-based resolution.
cmd/entire/cli/review/fix_findings_test.go	Add parsing tests for codex bolded severity bullets and severity title matching.
cmd/entire/cli/review/export_test.go	Export run-config mapping helper for external tests; remove multipicker export.
cmd/entire/cli/review/dump.go	Switch per-agent dump rendering to muted markdown palette.
cmd/entire/cli/review/cmd_test.go	Update dispatch tests for role-driven flow, recursion guard, flags behavior, and model/effort mapping.
cmd/entire/cli/review/banner.go	Add context transparency banner formatting helper.
cmd/entire/cli/review/banner_test.go	Add tests for context banner output and pluralization.
cmd/entire/cli/review_context.go	Return structured context result (prompt + counts + items) for staging/banner and prompt composition.
cmd/entire/cli/review_context_test.go	Update tests for new context result return type and counts.
cmd/entire/cli/review_bridge.go	Remove unused deps field wiring related to legacy picker behavior.
cmd/entire/cli/mdrender/mdrender.go	Add muted rendering mode and wire through writer helpers; refactor renderer to accept styles.
cmd/entire/cli/lifecycle.go	Treat any non-empty ENTIRE_REVIEW_SESSION as a review session marker for env adoption.
cmd/entire/cli/agent/skilldiscovery/scan.go	Introduce shared skill discovery scanners + version selection + frontmatter parsing.
cmd/entire/cli/agent/skilldiscovery/registry.go	Remove codex hardcoded /review; adjust install-hint strings for huh rendering.
cmd/entire/cli/agent/skilldiscovery/registry_test.go	Update assertions for codex curated built-ins now being empty.
cmd/entire/cli/agent/codex/reviewer.go	Pass skills verbatim, add model/effort flags, add rollout token tailer + per-turn token emission.
cmd/entire/cli/agent/codex/reviewer_test.go	Update tests for verbatim skill pass-through, flag mapping, and per-turn token emission.
cmd/entire/cli/agent/codex/review_tokens.go	Add rollout JSONL token tailer to emit live cumulative token totals.
cmd/entire/cli/agent/codex/review_tokens_test.go	Add tests for rollout token parsing and tailing appended lines.
cmd/entire/cli/agent/codex/discovery.go	Implement codex $name/$plugin:name skill discovery using shared scanners.
cmd/entire/cli/agent/codex/discovery_test.go	Add discovery tests for user, plugin cache, and superpowers layouts.
cmd/entire/cli/agent/claudecode/testdata/stream_with_deltas.jsonl	Add fixture for Claude stream-json usage snapshot vs final result usage.
cmd/entire/cli/agent/claudecode/reviewer.go	Emit input-only live Tokens on assistant envelopes and final totals on result.
cmd/entire/cli/agent/claudecode/reviewer_test.go	Update tests for new live-token behavior and fixture expectations.
cmd/entire/cli/agent/claudecode/discovery.go	Delegate discovery logic to shared skilldiscovery package scanners.
CLAUDE.md	Update docs for role-driven review flow, new commands, and removed legacy constructs.