What problem does it solve? Solo builders often commit weeks to ideas that were never worth building. This Skill runs a structured one-hour kill gate that forces a raw idea through six questions, two independent cold research searches, and an anonymised third-party judge, ending in a recorded verdict — build, kill, or park — instead of an open-ended conversation. ## Core Features & Use Cases - Six-question kill gate: Questions asked one at a time, with question zero classifying the idea (novelty, execution, or building-itself) to arm the right hard kills and brief the search agents. - Cold research subagents: Two /research subagents search whether the idea already exists and whether the hard part is tractable, deliberately kept unaware of the user's enthusiasm to avoid mirroring bias. - Three mechanical hard kills: No first version, already-exists-and-is-good, and the-hard-part-is-the-whole-project fire without weighing or appeal. - Anonymised pitch-judge: A case file with attribution stripped and verdict order randomised is judged independently to break ties or override false approvals. - Use Case: You have a side-project idea and no repo or spec. Run the gate; if it survives, the Skill creates a private repo, seeds issue #1 with the project map, and hands off to charting. If killed, the reason is recorded in a private ideas file for future reference. ## Quick Start Ask the assistant to run the pitch skill on your raw idea, for example: "I have an idea for a CLI tool that converts meeting notes into Jira tickets — pitch this and tell me if it is worth building."
