Interrogate a vague proposal
Identify the goal, users, scope, constraints, success measures, dependencies, risks, ownership, and assumptions that need answers.
Codex Grill Me Skill turns a vague idea into a decision-ready plan by asking focused questions, challenging unclear answers, tracing dependencies, and exposing assumptions before implementation begins. Download the skill for product ideas, architecture, workflows, specifications, and decisions that are still too loose to build safely.
task: complete requirements interrogation task
inspect:
- requirements and context
- existing standards
- failure and edge cases
verify: outputs + checks + handoffA good grilling session is not a long questionnaire. It follows the answer, opens one important decision branch at a time, and stops when the plan contains enough evidence to act.
This skill helps Codex inspect available context first, ask precise questions, recommend a default when useful, challenge contradictions, and maintain a visible record of decisions and unresolved risks.
Identify the goal, users, scope, constraints, success measures, dependencies, risks, ownership, and assumptions that need answers.
Ask one consequential question at a time, use each answer to choose the next branch, and avoid questions already resolved by available context.
Point out contradictions, unsupported claims, missing failure behavior, premature solutions, and tradeoffs being treated as free.
Summarize decisions, rationale, requirements, exclusions, open questions, acceptance checks, and the next useful step.
The steps keep context, implementation, and verification visible so the result can be reviewed and repeated.
Read the supplied plan, repository, notes, constraints, and prior decisions so the interview does not waste the user's time.
Choose the question whose answer changes the most downstream decisions and explain why it matters.
Ask follow-ups when an answer is vague, test edge cases and consequences, and record the decision before moving on.
Return a concise plan with agreed requirements, rationale, risks, exclusions, unresolved questions, and readiness status.
The workflow adjusts to the project, audience, tools, and risk while preserving the same quality standard.
Clarify users, jobs, scope, business rules, permissions, states, metrics, rollout, support, and failure behavior.
Examine load, data, trust boundaries, integrations, operations, migration, cost, recovery, and alternative designs.
Define triggers, actors, inputs, handoffs, exceptions, approvals, outputs, ownership, and measurements.
Turn a loose request into behaviors, constraints, interfaces, edge cases, testable outcomes, and non-goals.
Start with one defined outcome and provide the source material, constraints, and checks that matter.
These skill is designed for people who need dependable requirements interrogation work with a visible process.
Pressure-test ideas and turn incomplete intent into a plan a team can evaluate.
Expose missing technical decisions, operational risks, and unstated constraints before implementation.
Clarify users, scenarios, evidence, states, accessibility needs, and learning goals.
Create a shared decision record that reduces interpretation between planning and execution.
The skill can challenge and organize a plan, but it cannot supply business authority, user evidence, legal judgment, or stakeholder decisions that have not been provided.
Install the complete skill folder and add the project-specific context before beginning.
Keep SKILL.md with interview rules, decision tracking, challenge patterns, stopping conditions, and output format.
Use project scope when the interview should inspect repository context, or personal scope for reusable planning sessions.
Add the idea, plan, notes, constraints, known decisions, research, technical context, and intended audience.
State whether you want a short critical pass or a full decision-tree interview before receiving the final brief.
Practical answers about capabilities, limits, setup, and review.
No. It prioritizes one consequential branch at a time so each answer can shape the next question.
Its primary job is clarification. Implementation should begin only after the resulting plan is sufficiently complete and you explicitly request it.
Yes. When repository context is available, it should check what is already decided before asking questions.
It will challenge contradictions, vague claims, missing evidence, and unpriced tradeoffs while keeping the discussion focused on the decision.
It stops when critical branches are resolved, remaining uncertainty is visible, acceptance checks exist, and the next step is clear.
Clear context. Purposeful work. Relevant checks. A result others can understand.