Grill (MEL/SRHR)

A relentless interview that sharpens a MEL/SRHR design before any drafting starts. It is the MEL-aware sibling of the generic grilling skill: same one-question-at-a-time discipline, but it walks the branches that decide whether a theory of change, evaluation, indicator set, or proposal will survive scrutiny.

Lane — when this skill, not another

• grill-mel (this): sharpen or pressure-test a MEL/SRHR design before building it. No orchestration, no spec pipeline.
• /ann: run the full MEL orchestration (complexity classification, Vi spawns specialists, quality gate). Use when the task is to produce the deliverable, not just sharpen the thinking.
• brainstorming (superpowers): net-new builds headed into a code spec → writing-plans pipeline (ane_package features, new skills).
• grilling (mattpocock): generic, non-MEL plan sharpening. Model-invocable, so it can collide with this skill on a MEL design. The description above resolves the collision by naming the MEL lane explicitly; when the design is MEL, this skill wins.
• /grill-me (mattpocock): the same generic grill, but user-invoked only (disable-model-invocation: true). It never auto-fires, so it never competes for model selection. Ane calls it by name.
• /decision-debate: brings independent perspectives to Ane on a contested call she has not answered. This skill instead pulls from Ane on a design she already leans on.

If the user actually wants the deliverable produced, stop grilling and hand off to /ann or the right builder skill.

Method

Interview the user relentlessly until you reach a shared, defensible understanding of the design. Walk down each branch one at a time, resolving dependencies between decisions before moving on.