team-execution

Execute a plan with automatic multi-reviewer consensus workflow

Validate and configure tmux + Claude settings for team-execution agent teams

Optional reviewer for team-execution. Reviews specifications, issue templates, SKILL.md files, CLAUDE.md sections, and task descriptions for AI-consumability: context completeness, unambiguous acceptance criteria, example coverage, constraint explicitness, and machine-parseable structure. Triggered when plan contains: issue template, GitHub issue, task description, acceptance criteria, AI prompt, SKILL.md, CLAUDE.md, spec. NOT for: dumbing down content; general documentation quality (clarity-reviewer). Focus: making specs structured and explicit for AI consumption.

Optional reviewer for team-execution. Reviews API design, contract correctness, versioning strategy, error response consistency, idempotency, and SDK impact. Triggered when plan contains: API, endpoint, REST, OpenAPI, versioning, deprecation, SDK, contract, breaking change. NOT for: implementation security (security-reviewer's job); infrastructure concerns.

Base reviewer for team-execution. Reviews implementations for design pattern consistency, separation of concerns, dependency direction, convention adherence, and architecture documentation coverage. Always spawned — present for every plan execution regardless of plan type. Context loading strategy: searches for architecture decision records or architecture docs in the project (docs/adrs/, architecture-decisions/, docs/architecture/) — loads only relevant ones if found; falls back to codebase pattern analysis if not. NOT for: code quality specifics (code-quality-reviewer); security-specific concerns (security-reviewer's job); test coverage (testing-reviewer).

Optional reviewer for team-execution. Reviews documentation and specifications for structure/navigation, precision of language, completeness, understandability (right audience level), and actionability. Triggered when plan contains: docs, README, specification, guide, runbook, architecture doc, documentation. Primarily used for docs/specs plans and mixed plans with doc content. NOT for: copy editing or grammar (focuses on meaning and structure); code quality.

Optional reviewer for team-execution. Reviews code for duplication, complexity, pattern consistency, naming/abstraction quality, and error handling patterns. Triggered when plan contains: refactor, lint, patterns, DRY, SOLID, complexity, code smell, technical debt, abstraction. NOT for: security concerns (security-reviewer); test coverage (testing-reviewer); style/formatting (linter handles that).

Two-phase structured plan execution with automatic multi-reviewer consensus workflow. Phase A runs DURING plan mode: reads the plan, derives workers from plan phases, detects optional reviewers from keywords, embeds a ## Team Structure section, and calls ExitPlanMode itself. The plan is a single atomic artifact — implementation steps, team roster, and review protocol approved together in one unit. Phase B is the orchestration protocol that Claude follows directly after plan approval: TeamCreate fires immediately (ONLY permitted first action), then plan approval gates, parallel execution, max-3-iteration review cycle with 9/10 consensus threshold, and completion reporting. Pattern source: adapted from a structured team review cycle for code/plan execution.