/grill-me: you explained it perfectly. Your agent built something else.

Matt Pocock's README: "The most common failure mode in software development is misalignment." His fix, /grill-me, started as two sentences: "Interview me relentlessly about every aspect of this plan until we reach a shared understanding."

Source: "grill-me", mattpocock/skills on GitHub, first published Feb 25, 2026 (Matt Pocock) https://github.com/mattpocock/skills Also: Theo on X (Aug 14, 2026), Peter Steinberger on X (Oct 13, 2026), Anthropic's Claude Code best practices (code.claude.com/docs/en/best-practices), and Mu et al., ClarifyGPT, FSE 2024 (arxiv.org/abs/2310.10996): 70.96% to 80.80% Pass@1 for GPT-4 on MBPP-sanitized.

The share-button prompt, the agent and its two builds are dramatizations; every quote and number on screen is from the sources.

Does your agent ask you questions before it codes?