orc-grill

What problem does it solve? Vague feature requests like "I want notifications for merchants" cannot be planned or built directly, and asking clarifying questions during code analysis wastes expensive scan tokens. This Skill runs a dedicated pre-planning interview lane that turns one vague sentence into a written, settled idea before any code is touched. ## Core Features & Use Cases - Structured interview rounds: Builds a design tree of open questions and asks the whole ready frontier each round, with numbered questions and recommendations so answers can be given by number. - Self-service fact lookup: Resolves factual questions itself via wiki status, cached patterns, gotcha lists, and read-only recon dispatches instead of making the user recite their own codebase. - Decision tagging and persistence: Tags every settled decision as intent or constraint and writes a grill-context.md document at the project root that later lanes consume. - Multiple exits: Save the idea, hand it directly into /orc-analyze, persist constraints to the pact ledger, write it up as a document, or drop it. - Use Case: You type "something is wrong with our refunds" and, after a few rounds of questions plus automatic fact lookups, receive a grill-context.md recording what you actually want, what was ruled out, and which open questions need which instrument. ## Quick Start Ask the assistant to run /orc-grill with a one-sentence idea such as "I want notifications for merchants" and answer the numbered interview rounds until the idea is confirmed and saved.