Guide software implementation planning with file-based persistence and measurable success criteria.
npx skills add https://github.com/iliaal/whetstone --skill ia-planning
Please help me install this Agent Skill. Skill: ia-planning Source: https://github.com/iliaal/whetstone/tree/main/plugins/whetstone/skills/ia-planning Command: npx skills add https://github.com/iliaal/whetstone --skill ia-planning
This Skill prevents AI coding agents from jumping straight into implementation without deciding what “done” means, how to verify it, and what scope boundaries to respect.
.plan/ working state so key decisions and progress survive context resets.Ask the agent to produce an implementation plan for a feature that touches multiple files and has ambiguous scope, and ensure it first gates success criteria with a measurable verification step.
High-intent search queries and answers about installing and using this skill.
To create a verifiable implementation plan for a multi-file refactor, you must establish measurable success criteria, discover existing tests, and produce concrete, phase-sized tasks with verifiable acceptance checks using a file-persisted workflow.
AI coding agents lose context during complex software engineering tasks because conversation history is volatile. Writing plan artifacts to a persistent `.plan/` directory on disk ensures key decisions and progress survive context resets.
A goal quality gate in software implementation planning is a checkpoint that forces measurable success criteria and explicit stop or ask points before planning begins, preventing agents from jumping straight into implementation without clear scope boundaries.
To prevent AI agents from drifting from the original software architecture plan, use a workflow that requires deviation documentation and traceability from completed tasks, including recording commit SHAs for every phase-sized implementation step.
Yes, you can use test discovery to plan verification against existing project conventions. The planning workflow detects the current test landscape to ensure the multi-phase implementation plan aligns with your established testing frameworks.
Yes, you need a `.plan/` directory to structure multi-phase code changes. This file-persisted working state captures key architectural decisions and progress, treating conversation context as volatile memory while maintaining verifiable plan artifacts.
NeKon69Resolve the project root directory by searching upward for marker files.
kehwarExecute a plan phase and update the plan and PRD with outcomes.
JoNaYeonGenerate unit tests matching existing Jest, Pytest, or Go testing patterns.
andrew-t-jamesGuide Socratic questioning to produce execution-ready specifications with acceptance criteria.