Multi-agent review profiles for entire review

[dipree]

https://entire.io/gh/entireio/cli/trails/466

Adds configurable multi-agent review profiles to the hidden experimental entire review command.

User flow

1. Configure a profile

entire review --configure
entire review --configure general --local
entire review --configure security \
  --set-slot claude-code=opus \
  --set-slot codex=o4-mini \
  --set-judge claude-code=opus \
  --set-output trail \
  --set-task "Review auth and data-access risks."

Profiles live in .entire/settings.json for shared config or .entire/settings.local.json for local config. Scripted setup supports --set-agents, repeatable --set-slot agent[=model], repeatable --set-model agent=model, --set-judge agent[=model], --set-output local|trail, and --set-task.

2. Inspect and edit configuration

entire review --list
entire review --agents --profile general
entire review --models
entire review --models --agent claude-code
entire review --edit --profile general

3. Run a review

entire review general
entire review --profile general --base origin/main --prompt "Focus on regressions" --timeout 15m
entire review --profile general --agent codex --model o4-mini
entire attach --review <session-id>

Reviewers run in parallel. One final judge runs after them and emits the concise verdict. --timeout defaults to 10m per reviewer; --timeout 0 disables the bound. A single-reviewer run is available with --agent, and --model only applies with --agent.

4. Read results

entire review --findings

Local output writes the review manifest and browsable findings. --set-output trail also posts the final verdict to the branch Trail as guarded structured findings when possible, with markdown fallback.

Hardening included

• Start and wait failures flow through terminal reviewer markers.
• Reviewer timeouts are cause-marked and do not duplicate synthetic run errors.
• Trail posting preserves line anchors and single-line ranges.
• Trail output is validated at the posting boundary.

Verification

• go build ./...
• go test ./...
• go test -race ./cmd/entire/cli/review/
• golangci-lint run ./...

[cursor]

Cursor Bugbot has reviewed your changes and found 5 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 bc13a81. Configure here.}

[Copilot]

Pull request overview

This PR redesigns the experimental entire review command around named review profiles with fan-out to multiple worker agents and mandatory master adjudication (for multi-worker profiles), replacing the legacy per-agent review config and removing the old migration flow.

Changes:

• Introduces review_profiles + review_default_profile settings, including worker-level agent aliases and model hints, plus a profile master used for final adjudication.
• Updates entire review CLI surface (--profile, positional profile name, --configure, --prompt, and --agent targeting a single worker).
• Implements profile-native master synthesis via a new AgentSynthesisProvider, and threads --model into Claude/Codex/Gemini review runner argv.

Reviewed changes

Copilot reviewed 27 out of 27 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
docs/architecture/review-command.md	Updates architecture documentation for profile-based review and master adjudication.
cmd/entire/cli/settings/settings.go	Adds settings schema/types for review profiles and default profile; keeps legacy fields for parsing.
cmd/entire/cli/root.go	Updates root command comment to reflect profiles.
cmd/entire/cli/review/types/sink.go	Extends per-run metadata to include display name vs registry name and model.
cmd/entire/cli/review/types/reviewer.go	Extends RunConfig with profile/task/model context for profile-based prompts and runs.
cmd/entire/cli/review/synthesis_sink.go	Adds agent-backed master synthesis provider and auto/adjudication-mode wiring.
cmd/entire/cli/review/synthesis_prompt.go	Updates synthesis prompt to an adjudication-style “final report” with rules and sections.
cmd/entire/cli/review/run.go	Records worker display name vs registry agent name and model in RunSummary; forwards events under display name.
cmd/entire/cli/review/run_multi.go	Propagates agent registry name and model through multi-run state and summaries.
cmd/entire/cli/review/prompt.go	Composes worker prompts from skills + profile name + canonical task + per-agent/per-run instructions.
cmd/entire/cli/review/profile.go	New profile resolution/default-task logic, worker/master selection, and preference persistence helpers.
cmd/entire/cli/review/picker.go	Adds guided setup flow and profile-aware editing; introduces master selection and model picking.
cmd/entire/cli/review/migration.go	Removes legacy migration flow from project settings to clone preferences.
cmd/entire/cli/review/migration_test.go	Removes tests for the deleted migration flow.
cmd/entire/cli/review/marker_fallback.go	Updates terminology/docs for “no adapter yet” manual fallback flow.
cmd/entire/cli/review/manifest.go	Improves manifest/session matching for worker aliases and model hints.
cmd/entire/cli/review/fix.go	Updates fix-agent choice derivation to pull from profiles (including master) and dedupe aliases.
cmd/entire/cli/review/export_test.go	Updates test export wrapper for changed synthesis prompt signature.
cmd/entire/cli/review/cmd.go	Rebuilds entire review command flow: profile resolution, guided setup, fan-out, and master synthesis.
cmd/entire/cli/review/cmd_test.go	Updates command tests for profile fan-out behavior and --prompt flow.
cmd/entire/cli/review_context_test.go	Updates review-context test settings to the new profile schema.
cmd/entire/cli/review_bridge.go	Removes legacy synthesis provider wiring; keeps reviewer wiring to avoid import cycles.
cmd/entire/cli/labs.go	Updates labs/experimental command summary text.
cmd/entire/cli/integration_test/review_test.go	Updates integration tests to configure reviews via profiles.
cmd/entire/cli/agent/geminicli/reviewer.go	Threads model hint into Gemini review runner command.
cmd/entire/cli/agent/codex/reviewer.go	Threads model hint into Codex review runner argv while keeping stdin ordering.
cmd/entire/cli/agent/claudecode/reviewer.go	Threads model hint into Claude review runner argv.