Interrogate software plans one question at a time to resolve ambiguity before coding.
npx skills add https://github.com/berad217/human-training --skill grill-berad217
Please help me install this Agent Skill. Skill: grill Source: https://github.com/berad217/human-training/tree/main/skills-source/grill Command: npx skills add https://github.com/berad217/human-training --skill grill-berad217
Fuzzy project plans, creeping scope, and unclear done criteria lead to misalignment between users and AI agents, resulting in wasted effort building the wrong software features before the mismatch is discovered.
Invoke the grill skill at the start of your project planning session to walk through every unresolved decision, clarify ambiguous terminology, and lock down a crisp, bounded plan before handing off to implementation.
High-intent search queries and answers about installing and using this skill.
To prevent scope creep and pre-coding misalignment, you need to interrogate your software plan to resolve every ambiguous decision. This process clarifies fuzzy feature ideas and establishes clear done criteria before implementation begins, avoiding wasted effort on misbuilt features.
The best way to clarify ambiguous domain terms is to challenge overloaded or conflicting definitions and capture them in a shared CONTEXT.md glossary. This creates a consistent shared language between you and your AI agent, ensuring accurate feature implementation.
To resolve unaddressed edge cases in a vague feature plan, apply adversarial plan interrogation by asking one targeted, recommendation-backed question at a time. This walks through every unresolved decision branch, locking down a crisp, bounded plan before handoff.
Yes, an AI coding agent requires clear done criteria before implementation to avoid building incorrect software features. Without bounded, actionable plans, fuzzy project criteria result in wasted effort and mismatched outcomes discovered too late in the development cycle.
You should not use adversarial plan interrogation if your software plan already has strict, unambiguous decision branches and a fully established CONTEXT.md glossary. It is designed for vague feature ideas with unaddressed edge cases rather than fully bounded, actionable plans.
oaustegardIdentify ambiguous user requests and ask targeted clarifying questions.
FabienDehopreRefines domain terminology and records architectural decisions in CONTEXT.md and ADR documentation.
eai-toolsExtract domain terminology from project documentation into feature glossaries.
cropsggInterrogate task requirements to align human intent with implementation.
