gds-quick-dev

Implements code changes through a gated plan-implement-review workflow with spec files.

Updated Mar 14, 2026
One-click install
npx skills add https://github.com/ArchaonHW/MingGoRTS --skill gds-quick-dev-archaonhw
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: gds-quick-dev
Source: https://github.com/ArchaonHW/MingGoRTS/tree/main/.agents/skills/gds-quick-dev
Command: npx skills add https://github.com/ArchaonHW/MingGoRTS --skill gds-quick-dev-archaonhw

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Turning a user request directly into code often skips planning, review, and traceability, producing unreviewable or misaligned changes. This Skill enforces a disciplined step-file workflow that converts any intent—feature, bug fix, refactor, or story—into a hardened, reviewable artifact with a persistent spec file. ## Core Features & Use Cases - Step-file workflow: Sequential clarify → plan → implement → review → present steps with mandatory human checkpoints and no step skipping. - Spec-driven development: Generates a token-bounded spec (900–1600 tokens) with frozen intent, code map, tasks, and Given/When/Then acceptance criteria. - Adversarial review loop: Runs blind, edge-case, and acceptance-auditor subagent reviews, classifying findings into intent_gap, bad_spec, patch, defer, or reject with loopback handling. - Sprint status sync: Updates sprint-status.yaml for epic stories and compiles focused epic context files from planning artifacts. - Use Case: A developer asks to add a FlowField pathfinding tweak to the MingGoRTS gameplay layer; the Skill drafts a spec, gets approval, implements via subagent, runs three review passes, and commits with a suggested review order. ## Quick Start Ask the agent to implement a specific feature or bug fix, for example: implement a small change to the battle planner and walk me through the review.

Frequently Asked Questions about gds-quick-dev

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I implement a feature with a plan-first AI workflow?

Invoke the skill with your intent; it clarifies requirements, drafts a spec file with tasks and acceptance criteria, waits for your approval, then implements and reviews the change. You approve or edit the spec at a checkpoint before any code is written.

What is the difference between the one-shot and plan-code-review routes?

One-shot handles zero-blast-radius changes with clear intent and no architectural decisions, implementing directly with a single adversarial review. Plan-code-review covers everything else with full spec drafting, approval checkpoint, implementation, and three-reviewer analysis.

How does the adversarial review classify findings?

Findings are deduplicated and classified as intent_gap, bad_spec, patch, defer, or reject. Intent gaps and bad specs trigger loopbacks that revert code and re-derive from an amended spec, while patches are auto-fixed and defers go to a deferred-work file.

Does this workflow work without subagent support?

Yes. Runtimes without subagents (e.g., Copilot, Codex, local models) fall back to inline execution: the skill writes review prompt files and halts, asking you to run each review in a separate session and paste back findings.

What are the spec size limits and why?

Specs target 900–1600 tokens and a single user-facing goal. Below 900 risks ambiguity; above 1600 risks context rot in implementation agents. Oversized or multi-goal intents trigger a split-or-keep decision with the human.