gsd-by-glittercowboy

Name: gsd-by-glittercowboy
Author: oleg-schmidt

功能描述

Get Shit Done - Full project planning and execution workflow. Handles project initialization with deep context gathering, automated research, roadmap creation, phase planning, and execution with verification.

使用说明 (SKILL.md)

\x3Cobjective> GSD (Get Shit Done) provides a complete workflow for taking projects from idea to execution through systematic planning, research, and phase-based development.

Full workflow port from Claude Code - Includes:

Deep questioning and context gathering
Automated domain research (4 parallel researchers)
Requirements definition and scoping
Roadmap creation with phase structure
Phase planning with research and verification
Wave-based parallel execution
Goal-backward verification

This is the complete GSD system, not a simplified version. \x3C/objective>

\x3Cintake> What would you like to do?

Core workflow commands:

new-project - Initialize a new project with deep context gathering, research, requirements, and roadmap
plan-phase [N] - Create execution plans for a phase (with optional research)
execute-phase [N] - Execute all plans in a phase with wave-based parallelization
progress - Check project status and intelligently route to next action
debug [issue] - Systematic debugging with persistent state across context resets
quick - Execute ad-hoc tasks with GSD guarantees but skip optional agents
discuss-phase [N] - Gather context through adaptive questioning before planning
verify-work [N] - Validate built features through conversational UAT
map-codebase - Analyze existing codebase for brownfield projects
pause-work - Create handoff when pausing mid-phase
resume-work - Resume from previous session with full context
add-todo [desc] - Capture idea or task for later
check-todos [area] - List and work on pending todos
add-phase \x3Cdesc> - Add phase to end of milestone
insert-phase \x3Cafter> \x3Cdesc> - Insert urgent decimal phase
remove-phase \x3CN> - Remove future phase and renumber
new-milestone [name] - Start new milestone cycle
complete-milestone \x3Cver> - Archive milestone and tag
audit-milestone [ver] - Verify milestone completion
settings - Configure workflow toggles and model profile

Flags:

plan-phase [N] --research - Force re-research before planning
plan-phase [N] --skip-research - Skip research, plan directly
plan-phase [N] --gaps - Gap closure mode (after verification finds issues)
plan-phase [N] --skip-verify - Skip plan verification loop
execute-phase [N] --gaps-only - Execute only gap closure plans

Usage:

/gsd new-project - Start a new project
/gsd plan-phase 1 - Plan phase 1
/gsd execute-phase 1 - Execute phase 1
/gsd progress - Check where you are and what's next
/gsd debug "button doesn't work" - Start debugging session
/gsd quick - Quick ad-hoc task without full ceremony
Or just tell me what you want and I'll guide you through GSD

What GSD does:

Deep questioning - Understand what you're building through conversation
Research - 4 parallel researchers investigate domain (stack, features, architecture, pitfalls)
Requirements - Define v1 scope through feature selection
Roadmap - Derive phases from requirements (not imposed structure)
Phase planning - Create executable plans with tasks, dependencies, verification
Execution - Run plans in parallel waves with per-task commits
Verification - Check must_haves against actual codebase \x3C/intake>

\x3Crouting> Based on user input, route to appropriate workflow:

Intent	Workflow
"new project", "initialize", "start project"	workflows/new-project.md
"new-project" (explicit)	workflows/new-project.md
"plan phase", "plan-phase", "create plan"	workflows/plan-phase.md
"execute phase", "execute-phase", "start work"	workflows/execute-phase.md
"progress", "status", "where am I"	workflows/progress.md
"debug", "investigate", "bug", "issue"	workflows/debug.md
"quick", "quick task", "ad-hoc"	workflows/quick.md
"discuss phase", "discuss-phase", "context"	workflows/discuss-phase.md
"verify", "verify-work", "UAT", "test"	workflows/verify-work.md
"map codebase", "map-codebase", "analyze code"	workflows/map-codebase.md
"pause", "pause-work", "stop work"	workflows/pause-work.md
"resume", "resume-work", "continue"	workflows/resume-work.md
"add todo", "add-todo", "capture"	workflows/add-todo.md
"check todos", "check-todos", "todos", "list todos"	workflows/check-todos.md
"add phase", "add-phase"	workflows/add-phase.md
"insert phase", "insert-phase", "urgent phase"	workflows/insert-phase.md
"remove phase", "remove-phase", "delete phase"	workflows/remove-phase.md
"new milestone", "new-milestone", "next milestone"	workflows/new-milestone.md
"complete milestone", "complete-milestone", "archive"	workflows/complete-milestone.md
"audit milestone", "audit-milestone", "audit"	workflows/audit-milestone.md
"settings", "config", "configure"	workflows/settings.md

\x3C/routing>

\x3Carchitecture>

Workflow Files

Located in workflows/:

new-project.md - Full project initialization workflow
plan-phase.md - Phase planning with research and verification
execute-phase.md - Wave-based execution orchestrator
progress.md - Status check and intelligent routing to next action
debug.md - Systematic debugging with persistent state
quick.md - Ad-hoc tasks with GSD guarantees, skip optional agents
discuss-phase.md - Gather context through adaptive questioning
verify-work.md - Conversational UAT to validate built features
map-codebase.md - Parallel codebase analysis for brownfield projects
pause-work.md - Create handoff when pausing mid-phase
resume-work.md - Resume with full context restoration
add-todo.md - Capture ideas/tasks for later
check-todos.md - List and work on pending todos
add-phase.md - Add phase to end of milestone
insert-phase.md - Insert urgent decimal phase
remove-phase.md - Remove future phase and renumber
new-milestone.md - Start new milestone cycle
complete-milestone.md - Archive milestone and tag
audit-milestone.md - Verify milestone completion
settings.md - Configure workflow toggles

Agent Files

Located in agents/:

gsd-project-researcher.md - Research domain ecosystem (stack, features, architecture, pitfalls)
gsd-phase-researcher.md - Research how to implement a specific phase
gsd-research-synthesizer.md - Synthesize parallel research into cohesive SUMMARY.md
gsd-roadmapper.md - Create roadmap from requirements and research
gsd-planner.md - Create detailed execution plans for a phase
gsd-plan-checker.md - Verify plans will achieve phase goal before execution
gsd-executor.md - Execute a single plan with task-by-task commits
gsd-verifier.md - Verify phase goal achieved by checking must_haves against codebase
gsd-debugger.md - Investigate bugs using scientific method with persistent state
gsd-codebase-mapper.md - Analyze existing codebase for brownfield projects
gsd-integration-checker.md - Verify cross-phase integration and E2E flows

Reference Files

Located in references/:

questioning.md - Deep questioning techniques and context checklist
ui-brand.md - UI/UX principles and brand guidelines

Templates

Located in templates/:

project.md - PROJECT.md template
requirements.md - REQUIREMENTS.md template
research-project/ - Research output templates (STACK, FEATURES, ARCHITECTURE, PITFALLS, SUMMARY)

Workflow Pattern

GSD uses orchestrator + subagent pattern:

Orchestrator (workflow) - Stays in main context, spawns subagents, routes flow
Subagents (agents) - Fresh context, focused task, return structured result
Iteration - Verification loops (planner → checker → planner) until quality gates pass

This allows:

Lean orchestrator context (~15%)
Fresh context per subagent (100%)
Parallel execution (4 researchers, multiple plans in wave)
Verification before wasting execution time \x3C/architecture>

\x3Csuccess_criteria>

User can initialize new projects via /gsd new-project
Full workflow executes: questioning → research → requirements → roadmap
Phase planning includes research and verification loop
Phase execution uses wave-based parallelization
Verification checks must_haves against actual code
.planning/ directory structure created with all artifacts
Clear next steps provided at each stage \x3C/success_criteria>

安全使用建议

This skill appears internally consistent with a full project orchestration assistant, but it performs wide-ranging local file operations and git commits, spawns subagents, and can fetch external web content. Before installing or invoking: 1) back up your repository or run it in an isolated workspace; 2) review .planning/config.json (it controls whether docs are auto-committed) and set commit_docs=false if you want to avoid automatic git commits; 3) be prepared that the skill will create/move/delete files under .planning/, may remove REQUIREMENTS.md during milestone completion, and will spawn persistent debug sessions; 4) if you have sensitive data in the repo, avoid running WebFetch/WebSearch steps or provide only sanitized inputs. No credentials are requested by the skill itself.

功能分析

Type: OpenClaw Skill Name: gsd Version: 1.0.1 The OpenClaw AgentSkills skill bundle "gsd" (Get Shit Done) provides a comprehensive AI-driven development workflow. The analysis reveals a system designed for extensive interaction with the local filesystem, Git, external APIs (LLMs, cloud services), and the ability to modify code and configurations. **Key Observations and Capabilities:** 1. **Broad Tool Access:** Multiple agents (`gsd-codebase-mapper`, `gsd-debugger`, `gsd-executor`, `gsd-phase-researcher`, `gsd-planner`, `gsd-project-researcher`, `gsd-verifier`) are granted powerful tools including `Read`, `Write`, `Edit`, `Bash`, `Grep`, `Glob`, `WebSearch`, `WebFetch`, and `mcp__context7__*`. 2. **Code Modification and Execution:** The `gsd-executor` and `gsd-debugger` agents have `Edit` and `Bash` tools, allowing them to modify arbitrary files, write code, and execute arbitrary shell commands. The `gsd-executor` is explicitly instructed to commit changes to Git after each task. 3. **Network Access:** `WebSearch` and `WebFetch` tools are used by several agents (`gsd-debugger`, `gsd-executor`, `gsd-phase-researcher`, `gsd-project-researcher`, `gsd-planner`) for research and fetching resources. The `templates/phase-models-template.json` and `references/ccr-integration.md

能力评估

✓ Purpose & Capability

Name/description (project planning & execution) matches the delivered artifacts: many workflows for new-project, plan-phase, execute-phase, debug, design, discovery, etc. Required env vars/binaries/config paths are none, which is consistent with an instruction-only skill that operates over a project's .planning directory. Tools like WebSearch/WebFetch and git operations are reasonable for research and committing planning docs.

ℹ Instruction Scope

SKILL.md and workflow files instruct the agent to read/write project files under a .planning directory, run bash snippets, run git commits/tags, spawn subagents (sessions_spawn), and perform WebSearch/WebFetch for research. These actions are coherent with the stated purpose but are broad: they modify the repository (create/delete files, commit, tag), persist debug sessions, and perform network fetches. Users should understand it will perform repository-level changes and external searches if invoked.

✓ Install Mechanism

Instruction-only skill with no install spec and no downloaded code. No archives, brew/npm installs, or external URL downloads are used. This is the lowest-risk install model.

✓ Credentials

The skill declares no required environment variables, credentials, or privileged config paths. Workflows reference only local project files (e.g., .planning/*) and internal skill files. WebFetch/WebSearch usage implies network access but does not require user-provided secrets; this is proportionate to the research capabilities described.

ℹ Persistence & Privilege

always:false (normal). The skill instructs spawning persisted subagent sessions and writing persistent files under .planning (debug sessions, DESIGN-SYSTEM.md, DISCOVERY.md, etc.). Autonomous model invocation is allowed (platform default), so the skill could autonomously spawn subagents and perform repo edits when invoked. This is expected for an orchestrator skill but increases blast radius compared to read-only assistants.

版本历史

v1.0.1

- the skill was initially published under https://github.com/openclaw/skills/tree/main/skills/glittercowboy/gsd as there was no uploaded instance on clawhub.ai to trigger the security scan the skill was uploaded as is - no credits whatsoever goes to the uploader - Initial public release of the full GSD (Get Shit Done) project planning and execution workflow. - Provides complete end-to-end workflow: deep context gathering, automated 4-parallel researcher system, requirements definition, roadmap and phase planning, wave-based execution, and goal-backward verification. - Rich set of workflow commands for project lifecycle management: planning, execution, debugging, todos, milestones, and more—accessible via explicit commands or natural language. - Robust agent-based system: orchestrator coordinates dedicated agents for research, planning, execution, verification, and debugging. - Designed for both greenfield and brownfield projects; includes codebase mapping and adaptive questioning. - Includes workflow customization options and persistent context for smooth project handoffs and resumptions.

元数据

Slug gsd

版本 1.0.1

许可证 —

累计安装 8

当前安装数 8

历史版本数 1

常见问题