flowstate

Start a guided design session. Explores requirements through one-at-a-time questions, proposes approaches, and produces a validated design document. Use before any implementation work.

Capture learnings from the current work cycle. Documents what worked, what didn't, and how to prevent similar issues. Stores in docs/solutions/ for future sessions.

Systematic debugging with 6-step protocol: reproduce, isolate, diagnose, fix, verify, compound. 3-strike escalation.

Extended review with 14+ agents including conditional and language-specific reviewers. Use for large changes, architectural shifts, or when standard review isn't thorough enough.

Enhance an existing plan with parallel research agents per section. Adds best practices, edge cases, code examples, and learnings references. Does not rewrite — only enriches.

Re-inject Flowstate's core rules and workflow summary. Use when Claude drifts from the workflow or forgets the pipeline order.

Run 5 core review agents in parallel plus learnings researcher to catch issues before merge. Produces prioritized findings (P1/P2/P3) with todo files for tracking.

Learn how Flowstate works — shows the workflow pipeline, available commands, and getting-started guide.

Execute an implementation plan for Atlassian Forge apps with automated verification (yarn lint, tsc), structured tunnel verification, worktree isolation, subagent dispatch, and two-stage code review.

Execute an implementation plan task-by-task with strict TDD discipline, worktree isolation, subagent dispatch, and two-stage code review. Produces tested, committed, PR-ready code.

Transform a brainstorm or feature description into an actionable implementation plan with TDD-structured tasks, informed by past learnings and parallel research.

Searches external sources for industry standards, best practices, and real-world examples relevant to a feature or technical decision. Use when local context is insufficient or the work involves high-risk areas like security, payments, or external APIs.

Looks up framework and library documentation using Context7 MCP for accurate, version-specific API details and code examples. Use when implementation involves specific framework APIs, library methods, or version-dependent behavior.

Searches docs/solutions/ for relevant past solutions using grep-first filtering on YAML frontmatter. Use before implementing features or fixing problems to surface institutional knowledge and prevent repeated mistakes. This is the most critical agent for the compounding flywheel.

Scans codebase for existing patterns, conventions, and similar implementations. Use before planning or implementing features to understand the current codebase landscape.

Analyzes feature specifications to map user flows, identify missing error handling, surface edge cases, find ambiguities, and validate acceptance criteria. Use during planning to catch gaps before implementation begins.

Evaluates code changes for architectural integrity: component boundaries, dependency direction, separation of concerns, API stability, and design pattern consistency. Use when reviewing changes that affect system structure.

Detects design patterns, anti-patterns, naming inconsistencies, and code duplication. Use when reviewing code to ensure pattern consistency and catch common code smells.

Analyzes code for performance bottlenecks, algorithmic complexity, database query issues, caching gaps, memory problems, and scalability risks. Use when reviewing code changes or investigating slowness.

Performs security audits for vulnerabilities, injection flaws, auth/authz gaps, hardcoded secrets, and OWASP Top 10 compliance. Use when reviewing code changes before merge or deployment.

Reviews code for unnecessary complexity, over-engineering, dead code, and YAGNI violations. Use as a final review pass to ensure code is as simple and minimal as possible.

Use before any creative work — features, components, changes, or modifications. Guides structured design through collaborative dialogue before implementation. Triggers: "brainstorm", "design", "think through", "explore approaches", ambiguous requests, or requests with multiple valid interpretations.

Capture solved problems as categorized documentation with YAML frontmatter. Trigger words: "compound", "capture learning", "document solution", "that worked", "it's fixed", "working now", "problem solved", "that did it", "doc-fix"

Use for extended multi-agent review with 14+ agents including conditional and language-specific reviewers. Appropriate for large changes, architectural shifts, migrations, or when standard review is not thorough enough. Triggers: "deep review", "thorough review", "full review", "extended review", "review everything", "maximum coverage".

Use to review brainstorm or plan documents before proceeding to the next workflow step. Applies structured self-review for completeness, clarity, consistency, feasibility, and YAGNI.

Use when implementation is complete and all tests pass — guides branch completion by presenting structured options for merge, PR, or cleanup.

Use when performing multi-agent code review after implementing features, before merging, or when reviewing PRs. Runs 5 core review agents in parallel plus a learnings researcher. Produces prioritized findings (P1/P2/P3) with todo files for tracking and resolution. Triggers: "review", "code review", "check before merge", "review PR", "run reviewers", "catch issues".

Enhance an existing plan with parallel research agents per section. Adds best practices, edge cases, code examples, and learnings references. Does not rewrite -- only enriches. Triggers: "deepen", "enhance plan", "add research to plan", or after initial planning when more depth is needed.

Transform brainstorm outputs, feature descriptions, or improvement ideas into well-structured implementation plans with TDD-structured tasks. Orchestrates parallel research (local + conditional external), spec-flow analysis, and plan writing. Triggers: "plan", "implement", "build this", or after a brainstorm session when ready to move to implementation.

Use when receiving code review feedback, before implementing suggestions. Requires technical rigor and verification -- not performative agreement or blind implementation.

Use when requesting code review after implementing features, before merging, when stuck, or after fixing complex bugs. Structures what to provide reviewers and what output to expect.

Use when executing implementation plans with independent tasks — dispatches a fresh subagent per task with two-stage review (spec compliance then code quality) after each.

Use when encountering any bug, test failure, or unexpected behavior. Enforces a disciplined reproduce-isolate-diagnose-fix-verify cycle. No guessing allowed.

Use when implementing any feature or bugfix, before writing implementation code. Enforces strict test-first development — no production code without a failing test.

Wave-based parallel task execution using team agents with worktree isolation. Invoked by subagent-driven-development when plan has parallelizable tasks. Spawns up to 3 implementer teammates per wave, each in its own worktree. Reviews stay subagent-dispatched. Merges worktrees after each wave.

Create and manage git worktrees for isolated feature development — used at start of work phase after plan approval

Use when about to claim work is complete, fixed, or passing — requires running verification commands and confirming output before making any success claims. Evidence before assertions, always.

Structure implementation plans with TDD-structured tasks, bite-sized granularity, and learnings integration. Use when converting research/specs into an actionable plan document.

1 hook across 1 event