Assess task complexity and determine workflow type (feature, refactor, investigation, migration, or simple). Creates .workflow/ directory with assessment.json.

Compact workflow memory files to reduce context bloat. Summarizes patterns.md and gotchas.md, archives old entries. Use when memory files are large or before long implementation runs.

Analyze codebase for improvements across 6 dimensions: performance, security, quality, improvements, docs, ui-ux. Run a single mode or all at once.

Execute the next pending subtask from the implementation plan. Reads plan.json, respects phase dependencies, implements one subtask, runs self-critique and verification.

Autonomously manage the full development lifecycle: assess → spec → plan → implement → QA. Reasons about complexity, spawns specialist agents, passes context between phases, adapts to failures.

Create a phased implementation plan with subtasks, dependencies, and verification criteria. Reads spec.md if available, outputs plan.json.

Run quality assurance validation against the spec. Verifies subtask completion, runs tests, performs security review, checks pattern compliance, and generates a QA report with verdict.

Multi-specialist code review: security, logic, quality, and codebase fit analysis. Reviews uncommitted changes, PRs, or specific files with evidence-based findings.

Strategic feature planning: discover project purpose, research competitors, and generate a prioritized feature roadmap. Run individual stages or the full pipeline.

Write a grounded specification document based on codebase investigation. Reads assessment if available, investigates existing patterns, and outputs spec.md with requirements and acceptance criteria.

Show workflow progress for the current task or all tasks. Displays pipeline stage, phase completion, and next actions.

Assess task complexity and classify workflow type. Use when you need to determine if a task is simple, standard, or complex before deciding on an approach. Produces assessment.json.

Execute a single subtask from an implementation plan. Reads context, implements the change following existing patterns, runs mandatory self-critique, verifies, and updates plan status.

Deep codebase exploration specialist. Use when you need to understand existing patterns, conventions, and relevant code before designing or implementing a feature. Returns structured investigation findings.

Create a phased implementation plan with subtasks, dependencies, and verification criteria. Use when you have a spec and need to break it into an actionable plan.json.

Run quality assurance validation against a spec. Verifies subtask completion, runs tests, performs security review, checks pattern compliance, and generates qa_report.md with APPROVED/REJECTED verdict.

Multi-specialist code review across 4 dimensions: security, logic, quality, and codebase fit. Every finding requires code evidence. Use for reviewing changes before committing or merging.

Write a grounded specification document from accumulated context. Use when you have investigation findings and need to produce a formal spec.md with requirements, patterns, and acceptance criteria.