Agent Skills: Grill Me

Stress-test plans and designs through a one-question-at-a-time interview. Use when the user asks to 'grill me', pressure-test a repository strategy, module architecture, feature design, OpenSpec namespace spec, or proposed change. Walk each decision branch, surface dependencies, and recommend an answer for every question.

UncategorizedID: arisng/github-copilot-fc/grill-me

Install this agent skill to your local

pnpm dlx add-skill https://github.com/arisng/github-copilot-fc/tree/HEAD/skills/grill-me

Skill Files

Browse the full folder contents for grill-me.

Download Skill

Loading file tree…

skills/grill-me/SKILL.md

Skill Metadata

Name
grill-me
Description
"Stress-test plans and designs through a one-question-at-a-time interview. Use when the user asks to 'grill me', pressure-test a repository strategy, module architecture, feature design, OpenSpec namespace spec, or proposed change. Walk each decision branch, surface dependencies, and recommend an answer for every question."

Grill Me

Relentlessly interview the user until there is shared understanding across all important decision branches.

Supported Scopes

Use this workflow for any of these targets:

  • Repository or project level strategy
  • Module level architecture
  • Feature design inside a module
  • OpenSpec living spec namespace or a specific module capability inside a namespace
  • OpenSpec proposed change

Core Rules

  • Ask exactly one question at a time.
  • For each question, provide your recommended answer before waiting for user confirmation.
  • If a question can be answered from the codebase or docs, investigate first and present evidence-backed findings instead of asking blindly.
  • Keep drilling until each branch is resolved, explicitly deferred, or marked out of scope.
  • Do not move to implementation advice until the decision tree is stable.

Procedure

  1. Frame the scope and objective.
  • Confirm what is being stress-tested (repository, module, feature, namespace/capability, or change).
  • Define the target outcome (approved design, implementable plan, review checklist, or decision record).
  1. Build the initial decision tree.
  • Identify major branches relevant to the scope:
    • Goals and success criteria
    • Functional behavior and scenarios
    • Boundaries, ownership, and interfaces
    • Data and state concerns
    • Security, permissions, and compliance
    • Performance, reliability, and failure handling
    • Testing and verification strategy
    • Rollout, migration, and rollback
    • Open risks, assumptions, and unknowns
  • Order branches by risk and dependency so upstream decisions are resolved first.
  1. Run the interview loop.
  • Ask one high-impact question.
  • Include:
    • Why the question matters
    • Recommended answer
    • Impact of accepting that recommendation
  • Wait for the user answer before proceeding.
  1. Branch resolution logic.
  • If the user agrees: lock the decision and move to the next dependent branch.
  • If the user disagrees: capture the alternative, test consequences, and ask the next narrowing question.
  • If unknown: investigate code/spec/docs; if still unknown, mark as explicit risk with owner and trigger.
  • If blocked by external dependency: record defer condition and continue with independent branches.
  1. Evidence-first behavior.
  • Prefer codebase exploration when facts are discoverable.
  • Cite concrete artifacts (files, endpoints, contracts, specs, tasks) before escalating to user questions.
  • Convert discovered facts into sharper follow-up questions.
  1. Finish with a closure packet.
  • Provide:
    • Final decision map (resolved branches)
    • Open questions and blockers
    • Assumptions to validate later
    • Recommended next action sequence
  • Stop only when shared understanding is explicit.

Quality Bar

A successful grilling session produces:

  • No unresolved high-risk branch without an explicit defer reason
  • Clear rationale for each major decision
  • Traceable dependencies between decisions
  • A practical next-step plan that can be executed or reviewed immediately

Prompts You Can Use

  • "Grill me on this repository refactor plan."
  • "Grill me on the Session module permission model."
  • "Grill me on this feature design in Lifeline."
  • "Grill me on this OpenSpec platform namespace capability spec."
  • "Grill me on openspec change zoom-admin-dashboard-oauth-ui."
Grill Me Skill | Agent Skills