Pressure-tests a half-formed product or feature idea on desirability, strategy, and user experience — never on feasibility or implementation — until it is sharp enough to scope.
SkillScore77Maintained
Automatic
◆Product Management pack
Last verified 8/3/2026
What it changes
Passed 17 of 22 test cases with the skill, and 4 of 22 without it.
Without skill18%
With skill77%
Checks it passed that the model alone missed
The output declines to interrogate the feature idea.
The output directs the user to change-scope for scoping.
The output declines to execute change-scope automatically.
and 18 more in the full report
13 fixed4 passed either way5 still failing0 broke
Pass-rate lift
+59 pp
Tokens
+35%
Agent turns
0%
Response time
-46%
Model gemini-3.6-flashMeasured 8/3/2026v117 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
What's inside
The instructions your model reads when the skill loads.
SKILL.md110 linessingle file
Idea Grill
What this skill does
idea-grill takes a vague product or feature idea and interrogates it —
adversarially, on desirability and strategy — until it is sharp enough to
scope. It produces a transient readied-idea stub that change-scope consumes.
It is a pre-stage, not a pipeline stage: it writes no PRD, no durable artifact,
no why-link. It grills the idea; change-scope authors the PRD from the
sharpened result.
When this skill applies
A vague, half-formed feature or product idea that needs sharpening before it
can be scoped — this skill.
A well-formed idea whose problem, value, and shape are already clear — NOT
this skill; go straight to change-scope. Grilling is the front door for
vague ideas, never a mandatory gate.
A product-level question (strategy, positioning, the product's reason to
exist) — NOT this skill; that is vision-grill.
Anything about how an idea would be built — NOT this skill; feasibility is
change-spec's job.
The grill
Interrogate the idea adversarially. Your job is to find the holes, not to be
agreeable. Pursue, in no fixed order, until each has a real answer:
The problem. What can the user not do today? Is the problem real, or
assumed? Who actually has it — and how do you know?
The value. What is shipping this worth, and to whom? Why now?
The user experience. What is the user's journey through this? Where is
it awkward? What is the experience if it goes wrong, or if data is missing?
The smallest worthwhile version. What is the least that could ship and
still be worth shipping? What is gold-plating?
The shape. Is this one capability, or several? Does it extend an
existing capability, or create a new one? (You classify loosely — a sharper
classification is change-scope's job.)
The non-goals. What is this idea deliberately NOT? Name the boundary.