apsulli

You can plan goals into phased execution plans, validate plans for cycles and coverage, execute them with atomic commits and checkpoints, debug systematically with persistent state, review code for bugs and race conditions, map multi-language codebases into ARCHITECTURE.md and STACK.md, and verify work against specs with documented evidence.

They target software engineers and developers running structured, plan-driven development sessions. The skills suit practitioners who need dependency-aware planning, evidence-based verification, systematic debugging, codebase architecture analysis, and credit-efficient prompt generation for AI coding environments like Claude Code, Cursor, and Copilot.

The GSD Planner converts goals into phase plans with dependency graphs and execution waves. The Plan Checker validates GAPL-format plans for completeness and cycles. The Executor then runs the plan using atomic commits, checkpoint-driven progression, deviation handling, and state management, with the Verifier confirming results against specifications.

The Empirical Validation skill requires documented proof before marking work complete, rejecting unverified claims. The GSD Verifier validates implementations against spec requirements with empirical evidence and generates VERIFICATION.md findings, while the code-reviewer analyzes bugs, race conditions, and quality issues in the current session context.

Prompt-master generates surgical, credit-efficient prompts for AI tools including Claude, ChatGPT, Gemini, Cursor, Copilot, Midjourney, and Stable Diffusion. It triggers on requests to write, fix, improve, or decompose prompts, eliminates wasted tokens, prevents scope creep, and asks clarifying questions when intent is ambiguous.