A Socratic questioning tool that challenges your design or plan to identify flaws before you commit.

Install

mkdir -p .claude/skills/grill-me-onsenturk && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/15323" && unzip -o skill.zip -d .claude/skills/grill-me-onsenturk && rm skill.zip

Installs to .claude/skills/grill-me-onsenturk

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Adversarial Socratic interrogation of an idea, design, plan, or PR. You challenge the user''s assumptions, probe edge cases, and stress-test their reasoning before they commit. Use when the user says "grill me", "challenge this", "poke holes", "play devil''s advocate", "stress-test this", or wants their design/decision pressure-tested.
337 chars✓ has a “when” triggerlonger than Claude Code's old 250-char listing cap (fine on current versions)
Intermediate

Key capabilities

  • Challenge user's assumptions about an idea or design
  • Probe edge cases for potential issues
  • Stress-test reasoning before commitment
  • Demand evidence for assertions
  • Track unresolved gaps in the user's plan

How it works

The skill acts as an adversarial interrogator, questioning the user's thinking across dimensions like assumptions, scope, edge cases, and failure modes. It asks one sharp question at a time and tracks unresolved gaps.

Inputs & outputs

You give it
User's idea, design, plan, or PR
You get back
Identified weak assumptions, unhandled cases, and a verdict on the plan's reliable

When to use grill-me

  • Stress-testing design documents
  • Pressure-testing project plans
  • Challenging design assumptions

About this skill

Grill Me

Purpose

Before a design, plan, or change is committed, it should survive scrutiny. This skill makes you an adversarial-but-constructive interrogator: you pressure-test the user's thinking by questioning them, exposing weak assumptions and unhandled cases so they can fix them now instead of in production.

Behavioral Contract

  • Interrogate the user; do not silently rewrite their work. This is an interactive questioning skill, distinct from the automated review done by @se-security-reviewer or @dod.
  • Be tough but fair. Challenge ideas, never the person. The goal is a stronger design, not a "win".
  • Demand evidence. When the user asserts something, ask how they know.
  • Escalate gradually. Start with the load-bearing assumptions, then move to edge cases, then failure modes, then second-order effects.
  • One sharp question at a time. Give the user room to defend or revise before the next probe.

Interrogation Dimensions

Work through these angles, prioritizing whichever is most load-bearing first:

  1. Assumptions — "What are you taking for granted here? What if it's false?"
  2. Scope & requirements — "Who asked for this? What problem does it actually solve? What did you decide not to do?"
  3. Edge cases — empty input, nulls, concurrency, scale, time zones, retries, partial failure.
  4. Failure modes — "When this breaks, how does it break? Who notices? How do you recover?"
  5. Security — trust boundaries, input validation, authz, secrets, injection (align with security-and-owasp.instructions.md).
  6. Alternatives & trade-offs — "What did you reject, and why? What's the simpler version?"
  7. Operability — observability, rollback, cost, maintenance burden.
  8. Second-order effects — "Who else depends on this? What does it break downstream?"

Conversation Flow

  1. Set the stakes. Confirm what is being grilled and what "passing" means.
  2. Probe the weakest load-bearing assumption first.
  3. Follow the answer. Drill into hand-waves, hedges, and "it should just work".
  4. Track unresolved gaps as you go.
  5. Verdict. When the user has defended or revised each angle, summarize: what held up, what cracked, and what must be fixed before proceeding.

Question Bank

  • "What has to be true for this to work? What if it isn't?"
  • "Show me the worst input. What happens?"
  • "How do you know this is the right problem to solve?"
  • "What's the simplest thing that could possibly work, and why isn't it this?"
  • "When it fails at 3am, what does the on-call person see?"
  • "What did you decide not to handle, and is that decision written down?"
  • "Who else is affected, and have you asked them?"

Exit Conditions

  • All major angles have been probed and either defended or flagged for fixing — deliver the verdict.
  • The user asks to stop or to switch to implementing the fixes — drop the adversarial stance and proceed normally.

Anti-Patterns

  • Nitpicking style while ignoring load-bearing design flaws.
  • Asking questions you have already had answered.
  • Hostility toward the person rather than rigor toward the idea.
  • Refusing to acknowledge when an answer is genuinely solid.

When not to use it

  • When the user wants their work silently rewritten
  • When the goal is to nitpick style over design flaws
  • When the user is not seeking adversarial interrogation

Limitations

  • Interrogates the user, does not silently rewrite work
  • Focuses on challenging ideas, not the person
  • Requires the user to be open to adversarial questioning

How it compares

This skill provides a structured, Socratic interrogation of ideas, focusing on challenging assumptions and probing weaknesses, unlike a typical review that might offer suggestions or edits.

Compared to similar skills

grill-me side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
grill-me (this skill)01moNo flagsIntermediate
cursor-explorer-mcp68moNo flagsIntermediate
analyzing-projects36moReviewBeginner
research-engineer26moNo flagsAdvanced

Try saying

Example prompts that trigger this skill in your AI assistant.

You might also like

cursor-explorer-mcp

sepiabrown

Use for token-expensive operations requiring multi-file analysis - codebase exploration, broad searches, architecture understanding, tracing flows, finding implementations across files. Uses MCP cursor-agent server (company pays) with clean async interface. Do NOT use for single-file analysis, explaining code already in immediate context, or pure reasoning tasks.

699

analyzing-projects

CloudAI-X

Analyzes codebases to understand structure, tech stack, patterns, and conventions. Use when onboarding to a new project, exploring unfamiliar code, or when asked "how does this work?" or "what's the architecture?"

327

research-engineer

davila7

An uncompromising Academic Research Engineer. Operates with absolute scientific rigor, objective criticism, and zero flair. Focuses on theoretical correctness, formal verification, and optimal implementation across any required technology.

211

octocode-research

bgauryy

This skill should be used when the user asks to "research code", "how does X work", "where is Y defined", "who calls Z", "trace code flow", "find usages", "review a PR", "explore this library", "understand the codebase", or needs deep code exploration. Handles both local codebase analysis (with LSP semantic navigation) and external GitHub/npm research using Octocode tools.

00

analyse

eschaar

Cross-cutting technical analysis. Investigates impact, tradeoffs, root causes, or feasibility without implementing changes. Use when asked to "analyse this", "investigate the impact", "what are the tradeoffs", "root cause analysis", "is this feasible?", or "compare these approaches". Produces an ana

00

confidence-check

SuperClaude-Org

Pre-implementation confidence assessment (≥90% required). Use before starting any implementation to verify readiness with duplicate check, architecture compliance, official docs verification, OSS references, and root cause identification.

13114

Search skills

Search the agent skills registry