Planning & Specification Workflows

Architecture consultation for Conductor orchestrator. Makes decisions about system design, patterns, component boundaries, and technical architecture. Can approve architectural choices within established patterns. Escalates novel patterns or breaking changes to user.

Simulate a 5-member expert board deliberation for major decisions. Use when evaluating plans, architecture choices, feature designs, or any decision requiring multi-perspective expert analysis. Triggers: 'board review', 'get expert opinions', 'board meeting', 'director evaluation', 'consensus review'.

Master coordinator for the Evaluate-Loop workflow v3. Supports GOAL-DRIVEN entry, PARALLEL execution via worker agents, BOARD OF DIRECTORS deliberation, and message bus coordination. Dispatches specialized workers dynamically, monitors via message bus, aggregates results. Uses metadata.json v3 for parallel state tracking. Use when: '/go ', '/conductor implement', 'start track', 'run the loop', 'orchestrate', 'automate track'.

Session start context priming -- loads relevant memories from Pensyve at the beginning of a session to provide continuity across sessions. Use when switching projects or needing historical context.

Technical architecture review agent for execution plans. Uses cto-advisor skill to evaluate technical decisions, architecture patterns, tech debt implications, and technology choices in plan.md. Triggered by: 'review technical plan', 'cto review', 'architecture review', or automatically during Step 2 (Evaluate Plan) for tracks involving architecture decisions, integrations, or infrastructure changes.

Specialized business logic evaluator for the Evaluate-Loop. Use this for evaluating tracks that implement core product logic — pipelines, dependency resolution, state machines, pricing/tier enforcement, packaging. Checks feature correctness against product rules, edge cases, state transitions, data flow, and user journey completeness. Dispatched by loop-execution-evaluator when track type is 'business-logic', 'generator', or 'core-feature'. Triggered by: 'evaluate logic', 'test business rules', 'verify business rules', 'check feature'.

Specialized code quality evaluator for the Evaluate-Loop. Use this for evaluating code implementation tracks where the deliverable is functional code — features, API routes, state management, utilities. Checks build integrity, type safety, code patterns, error handling, dead code, imports, test coverage, and naming conventions. Dispatched by loop-execution-evaluator when track type is 'feature', 'refactor', or 'infrastructure'. Triggered by: 'evaluate code', 'code review', 'quality check', 'build check'.

Specialized integration evaluator for the Evaluate-Loop. Use this for evaluating tracks that integrate external services — Supabase auth/DB, Stripe payments, Gemini API, third-party APIs. Checks API contracts, auth flows, data persistence, error recovery, environment config, and end-to-end flow integrity. Dispatched by loop-execution-evaluator when track type is 'integration', 'auth', 'payments', or 'api'. Triggered by: 'evaluate integration', 'test auth flow', 'check API', 'verify payments'.

Specialized UI/UX evaluator for the Evaluate-Loop. Use this for evaluating UI shell tracks, design system tracks, screen implementation tracks, or any track where the primary deliverable is visual/interactive UI. Checks design system adherence, visual consistency, layout structure, responsive behavior, component states, animations, accessibility baseline, and usability check (copy quality). Dispatched by loop-execution-evaluator when track type is 'ui', 'design-system', or 'screens'. Triggered by: 'evaluate UI', 'UI review', 'design review', 'visual audit'.

Opens the running app in a browser and verifies recent UI changes actually work. Use whenever the user wants a quick smoke test or sanity check of recent work, or says "go", "open in browser", "check in browser", "test your work", "make sure it works", "smoke test", "verify", "did it actually work", "make sure the form/page works", "check the form submits", "works on mobile" — even when they don't explicitly ask for browser testing. Also activates implicitly when the user appends "...and make sure it works" to a UI request. Skips design critique; for that, use go-ui.

AI Personal Knowledge Manager — Zettelkasten, Obsidian, PARA, Note-Taking System, Second Brain…

Evaluate-Loop Step 4: EVALUATE EXECUTION. This is the dispatcher agent — it determines the track type and invokes the correct specialized evaluator. Does NOT run a generic checklist. Instead dispatches to: eval-ui-ux (screens/design), eval-code-quality (features/infrastructure), eval-integration (APIs/auth/payments), eval-business-logic (generator/rules/state). Triggered by: 'evaluate execution', 'review implementation', 'check build', '/phase-review'. Always runs after loop-executor.

Evaluate-Loop Step 3: EXECUTE. Use this agent to implement tasks from a verified plan. Works through plan.md tasks sequentially, writes code, updates plan.md after every task, and commits at checkpoints. Uses TDD where applicable. Triggered by: 'execute plan', 'implement track', 'build feature', '/conductor implement' (execution phase). Only runs after plan has passed evaluation.

Evaluate-Loop Step 5: FIX. Use this agent when an evaluation (plan or execution) returns FAIL. Takes the evaluator's fix list, creates specific fix tasks in plan.md, executes the fixes, and triggers re-evaluation. Handles the loop-back mechanism of the Evaluate-Loop. Triggered by: evaluation FAIL verdict, 'fix issues', 'address evaluation failures'.

Evaluate-Loop Step 2: EVALUATE PLAN. Use this agent to verify an execution plan before any code is written. Checks scope alignment, overlap with completed work, DAG validity, dependency correctness, task clarity, and invokes Board of Directors for major tracks. Outputs PASS/FAIL verdict. Triggered by: 'evaluate plan', 'review plan', 'check plan before executing'. Always runs after loop-planner and before loop-executor.

Evaluate-Loop Step 1: PLAN. Use this agent when starting a new track or feature to create a detailed execution plan. Reads spec.md, loads project context, and produces a phased plan.md with specific tasks, acceptance criteria, and dependencies. Triggered by: 'plan feature', 'create plan', 'start track', '/conductor implement' (planning phase).

File-based message queue for inter-agent coordination. Used by workers AND board directors to communicate. Provides: progress updates, task completion signals, file locking, board deliberation. Core infrastructure for parallel execution.

Harshly critique strategic documents (validation reports, competitive analyses, business plans) to identify blind spots, challenge assumptions, expose weaknesses, and generate Conductor-aligned action plans. Use when you need brutal honesty, "red team" analysis, or want to stress-test your thinking before execution. Applies multiple analytical frameworks (pre-mortem, assumption hunting, competitive gaps, market dynamics) and synthesizes findings into executive summaries with specific, actionable recommendations mapped to Conductor context artifacts.

Product consultation for Conductor orchestrator. Clarifies scope, interprets ambiguous requirements, prioritizes features within a track. Can make scope interpretations that don't change deliverables. Escalates scope expansions or feature changes to user.

Quality assurance consultation for Conductor orchestrator. Sets test coverage requirements, validates quality gates, determines testing strategy. Can adjust coverage thresholds within ranges. Escalates skipping critical tests to user.

Runs after every track completion to extract learnings. Analyzes what worked, what failed, and what patterns emerged. Updates conductor/knowledge/patterns.md with new solutions and errors.json with new error patterns. Proposes skill updates if workflow improvements are identified. Triggered automatically by orchestrator after COMPLETE step.

Technical implementation consultation for Conductor orchestrator. Advises on implementation approach, dependency choices, and coding patterns. Can approve dependencies under size threshold. Escalates large dependencies or breaking tooling changes to user.

Manage Conductor tracks, phases, and tasks. Use when working with track status, updating task markers, or navigating between tracks. Enforces the Evaluate-Loop workflow.

把工作拆成安全、独立、可并行的子任务,并协调 subagent 避免重复劳动和冲突修改。适用于 Claude Code 通过清晰 delegation…