Menu
Initialize a new gsd-wired project with guided questioning. Use when starting a new project.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | init |
| description | Initialize a new gsd-wired project with guided questioning. Use when starting a new project. |
| disable-model-invocation | true |
| argument-hint | [full|quick|pr] |
Initialize a gsd-wired project using $ARGUMENTS mode (default: full).
You are a builder-partner helping set up a new project. Ask questions one at a time, wait for each answer before asking the next. Never batch multiple questions. Use GSD-familiar language (phases, plans, waves). Never mention beads, bead IDs, or graph structures to the developer.
When mode is "full" or no mode is specified, ask these questions in this exact order, one at a time, waiting for a response before asking the next:
When mode is "quick", ask only these three essential questions:
When mode is "pr", this is for importing an existing PR or issue into a gsd-wired project for review or integration work. Ask in this order:
Once you have collected all answers for the chosen mode, call the init_project MCP tool with the collected context as a single JSON object. Map each answer to the corresponding field:
project_name: Derive a concise project name from the "what" answerwhat: Direct from the first answer (what are you building)why: Direct from the second answer (motivation)who: Direct from the third answer (target users) — use empty string for quick modedone_criteria: Direct from the "what does done/success look like" answertech_stack: Direct from the tech stack answer — use empty string for quick modeconstraints: Direct from the constraints answer — use empty string for quick/pr moderisks: Direct from the risks answer — use empty string for quick/pr modemode: The init mode used ("full", "quick", or "pr")pr_url: The PR or issue URL if in pr mode — omit or use empty string for other modesAfter the init_project tool returns successfully, display:
Wait for the developer's response. If they say nothing, auto-proceed after 30 seconds and suggest next steps: running /gsd-wired:status to see the project dashboard, then planning the first phase.
Creates a new Cobra CLI command following the project's New*Cmd() pattern in internal/cli/. Handles command registration in root.go, flag setup, output rendering, and test scaffolding. Use when user says 'add command', 'new subcommand', 'create CLI command', or adds files to internal/cli/. Do NOT use for MCP tools or hooks.
Creates a new hook event handler in internal/hook/ following the handle*() dispatcher pattern. Adds event constant to events.go, handler function, dispatcher case, and tests using fake bd binary. Use when user says 'add hook', 'new hook event', 'hook handler', or modifies internal/hook/. Do NOT use for MCP tools, CLI commands, or graph queries.
Creates a new MCP tool following the handle*() pattern in internal/mcp/. Registers in registerTools(), adds args struct, result struct, handler function, and test using connectInProcess(). Use when user says 'add MCP tool', 'new tool', 'expose to Claude', or adds files to internal/mcp/. Do NOT use for CLI commands (internal/cli/), graph client methods, or Cobra commands.
Build modular CLI applications with Cobra framework. Use when structuring CLI commands, implementing modular command architecture, handling flags and arguments, or when user mentions Cobra, CLI modularity, command registration, or spf13/cobra.
Implements beads graph operations in internal/graph/ using the bd CLI wrapper pattern. Covers bead CRUD, label queries, tier classification, batch writes, and index management. Use when user says 'graph operation', 'bead query', 'add graph method', 'graph client', or modifies internal/graph/. Do NOT use for direct bd CLI usage, MCP tool handlers, or CLI command implementations in internal/cli/.
Writes tests using the fake_bd test binary pattern from internal/graph/testdata/fake_bd/. Covers building fake binary, setting FAKE_BD_* env vars for canned responses, capture files for arg/env verification, and serverState/hookState setup. Use when user says 'write tests', 'add test', 'test coverage', or needs to test graph/mcp/hook code that calls bd. Do NOT use for tests that don't invoke the bd CLI binary (e.g., pure utility functions, config parsing, CLI flag tests).