sgai

Interactive idea refinement using Socratic method to develop fully-formed designs. When your human partner says "I've got an idea", "Let's make/build/create", "I want to implement/add", "What if we". When starting design for complex feature. Before writing implementation plans. When idea needs refinement and exploration. ACTIVATE THIS AUTOMATICALLY when your human partner describes a feature or project idea - don't wait for /brainstorm command.

Use when debugging a web UI bug or validating a multi-step browser flow. Start or reuse background servers through run-long-running-processes-in-tmux using the deterministic current-directory-plus-purpose session name, not generic tmux names or kill-before-create setup. Prevent repo-root Playwright artifact leaks by resolving the retrospective screenshots directory once and prefixing every saved filename with it.

Use when Go code starts external processes, detaches process groups, or needs cancellation and interrupt handling with os/exec.

Go code review checklist based on official Go style guides. When reviewing Go code for style, idioms, and best practices

Use when writing or changing tests, adding mocks, tempted to add test-only methods to production code, or considering absence tests for deleted helpers, deleted client members, exported API shape, or removed routes. Prevents testing mock behavior, production pollution with test-only methods, mocking without understanding dependencies, and memorial tests whose sole purpose is to prove deleted internal/client surface stays gone while preserving tests for current observable behavior and current public/server contracts such as 404 removed-route checks.

Automatically scans GOAL.md for unchecked items, provides completion status summary, and enables coordinator to mark items as complete. When coordinator needs to verify project completion status or before marking work as complete. When coordinator needs to mark completed items in GOAL.md. Symptoms - manually going through GOAL.md line by line to check task completion, needing quick summary of pending vs completed tasks by category, and verifying all requirements are met before finalizing work.

Use when you need a detached tmux process. Derive one deterministic session name from the absolute current working directory plus a purpose suffix, reuse the matching owned session before creating it, and only kill that session during explicit cleanup or when you verified it is broken.

You MUST USE THIS SKILL TO COMMUNICATE TO THE ENVIRONMENT WHAT YOU ARE DOING, WHERE YOU ARE GOING, AND WHETHER YOU ARE BLOCKED (IE NEED TO SEND THE HUMAN PARTNER A MESSAGE)

Entry point for STPA (System Theoretic Process Analysis) hazard analysis. Guides through all 4 steps sequentially. When starting a new STPA analysis session. When the human partner mentions safety analysis, hazard analysis, or risk assessment. When analyzing control systems for potential failures.

Interactive wizard to compose valid GOAL.md files for SGAI with step-by-step guidance through 7 phases. When your human partner wants to create a new GOAL.md file, start a new SGAI project, configure agents for a software factory, or asks "help me set up GOAL.md", "I want to configure agents", "let's start a new project with SGAI".

Post-completion factory improvement analysis. Guides the retrospective agent through artifact discovery, pattern analysis, suggestion generation, and presenting proposed changes for approval. Use when the retrospective agent starts its post-completion phase. Treat `sgai/` overlay changes and `AGENTS.md` changes as co-equal categories when evidence supports them, prefer `sgai/` for process and tooling fixes, and report separate Skills, Agent Prompts, Snippets, and `AGENTS.md` buckets including empty buckets.

TDD for process documentation - test with subagents before writing, iterate until bulletproof. When you discover a technique, pattern, or tool worth documenting for reuse. When editing existing skills. When asked to modify skill documentation. When you've written a skill and need to verify it works before deploying. When skills fail to help agents, when documentation is unclear, when new patterns emerge that need standardization.

Use when starting any conversation - establishes mandatory workflows for finding and using skills, including using Skill tool before announcing usage, following brainstorming before coding, and creating TodoWrite todos for checklists

Guide for creating custom opencode agents with proper configuration, tools, permissions, and reasoning for specific tasks. When creating new opencode agents for security auditing, code review, planning, documentation, or specialized workflows; when built-in agents don't suffice; when configuring agents with restricted tools or custom prompts

Problem-Solving Dispatch; Dispatch to the right problem-solving technique based on how you're stuck; Stuck on a problem. Conventional approaches not working. Need to pick the right problem-solving technique. Not sure which skill applies.

Use when completing tasks, implementing major features, or before merging to verify work meets requirements - dispatches skills:code-reviewer subagent to review implementation against plan or requirements before proceeding

Use when creating or editing skills, before deployment, to verify they work under pressure and resist rationalization - applies RED-GREEN-REFACTOR cycle to process documentation by running baseline without skill, writing to address failures, iterating to close loopholes

Build AI agents using prompt-native architecture where features are defined in prompts, not code. Use when creating autonomous agents, designing MCP servers, implementing self-modifying systems, or adopting the "trust the agent's intelligence" philosophy.

Use when facing 3+ independent failures that can be investigated without shared state or dependencies - dispatches multiple Workbenches agents to investigate and fix independent problems concurrently

Create and setup a new OpenAI Agents SDK application with interactive guidance for language choice, agent type selection (Basic, Voice, Realtime), project setup, and automatic verification.

Create and setup a new Claude Agent SDK application with interactive guidance for language choice, project setup, and automatic verification.

Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation