grilling
/grilling
Use when a plan, design or scope needs stress-testing before code exists, including "grill me", "pressure-test this", "help me think it through", "should we", "how should we" and any change request whose decisions are still open. Runs a design-tree interview in rounds, one unblocked question set at a time, each with a recommended answer, until the frontier is empty and the user confirms. Read-only research and questions only; never write code, config or git state. Not for lookups, settled plans or one-line fixes.
- Author
- BaSui01 ↗
- Repository
- snow-app-skills ↗
- Licence
- MIT
- Last repository change
- 11.11.2026
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 11.11.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Use when a plan, design or scope needs stress-testing before code exists, including "grill me", "pressure-test this", "help me think it through", "should we", "how should we" and any change request whose decisions are still open. Runs a design-tree interview in rounds, one unblocked question set at a time, each with a recommended answer, until the frontier is empty and the user confirms. Read-only research and questions only; never write code, config or git state. Not for lookups, settled plans or one-line fixes.
Triggers from the description
Original author description
Use when a plan, design or scope needs stress-testing before code exists, including "grill me", "pressure-test this", "help me think it through", "should we", "how should we" and any change request whose decisions are still open. Runs a design-tree interview in rounds, one unblocked question set at a time, each with a recommended answer, until the frontier is empty and the user confirms. Read-only research and questions only; never write code, config or git state. Not for lookups, settled plans or one-line fixes.
How it works
- When not to use
- Hard boundaries
- 1. Build the tree
- 2. Find facts (never ask the user)
- 3. Ask the whole frontier each round
- 4. Record
- 5. Converge
SKILL.md preview
--- name: grilling description: Use when a plan, design or scope needs stress-testing before code exists, including "grill me", "pressure-test this", "help me think it through", "should we", "how should we" and any change request whose decisions are still open. Runs a design-tree interview in rounds, one unblocked question set at a time, each with a recommended answer, until the frontier is empty and the user confirms. Read-only research and questions only; never write code, config or git state. Not for lookups, settled plans or one-line fixes. allowed-tools: - filesystem-read - grep-search - grep-find - codebase-search - lsp-symbols - lsp-references - lsp-goto - lsp-diagnostics - todo-todo-manage - memory-manage - user-interaction-askUserQuestion - sub-agents-activate - sub-agents-continue - sub-agents-getResult - config-manage - skills-skill-execute --- # Grilling Goal: grind the plan until nothing is silently assumed. Done means every branch of the design tree has been visited, the user has decided each open point, and they explicitly confirm the shared understanding. ## When not to use - The answer is a fact or a file lookup, not a decision: look it up instead. - The decision is already settled and the user wants implementation: use `plan-execute`. - The task is a one-line fix with no design branch: just do it. ## Hard boundaries - Interview and research only. No code, config, database or Git changes, no side-effecting commands. Implementation starts only after the user confirms. - Facts are yours to find; decisions belong to the user. Never ask for anything you can look up. - Reply in the user's language. No decorative headings, no recap of the process. ## 1. Build the tree - Restate the target and known constraints in one sentence; if the target is vague, ask one scoping question. - Design tree: each decision is a node; its children are the decisions it unlocks. - Run `memory-search` first for earlier decisions and pitfalls; do not re-ask settled points. ## 2. Find facts (never ask the user) Pick the narrowest tool per gap and parallelize when possible: - Project code, symbols, call chains, contracts: `lsp-*`, `filesystem-read`; use `grep-search` only for literal text such as logs, config and comments. - Application settings truth: `config-get` / `config-list` (identify the right scope; do not guess files). - Plugin or MCP tool names and parameters: `toolSearch-discover`. - In-repo data flow, table structure, existing implementations: sub-agent `dev-research`. - External facts, version differences, official docs: sub-agent `agent_search`; require inline citations. - Genuinely complex trade-offs with multiple viable options: sub-agent `dev-plan`. Research is an unsettled prerequisite: only questions downstream of it wait. Ask the rest of the frontier now; do not stall. ## 3. Ask the whole frontier each round Frontier = decisions whose prerequisites are settled and that you can ask now without guessing. - Ask with `user-interaction-askUserQuestion`, one call per question, all questions of the round in one batch, and that batch contains nothing but the questions (never mix in research or file reads). - Give at least two options per question; the first is "Recommended: <full recommended answer>", the rest are concrete alternatives. The user can still add free text. - Never substitute a Q1/Q2 text list in the message for the interactive tool. - Wait for every answer before computing the next frontier; questions that depend on an open item belong to the next round. - If the user selects mutually exclusive options, ask exactly one follow-up on that point; do not pick for them. ## 4. Record - After each settled batch, update `todo-todo-manage` in the same batch as real work: decided items, open questions, questions blocked on research. This survives context compaction. - Report only deltas each round: newly settled decisions, assumptions that were overturned, the next frontier. - When the user makes a call that will affect later work, save it with `memory-save` (kind=decision): conclusion, rationale, alternatives dropped. ## 5. Converge The frontier is empty when every branch has been visited and nothing is silently assumed. Close with four items only: settled decisions, remaining risks and costs, dropped alternatives and why, suggested next verification. Before asking for the go-ahead, load `decision-notes` and write one `proposed` note: Problem, Proposal, Alternatives considered, Acceptance criteria, Risks. One note per decision area, not one per question. Skip the note only when the user explicitly says the decisions do not need to be kept, or when the change is mechanical. Settle cross-cutting terms in `GLOSSARY.md` only when a term is overloaded or used differently across modules: a glossary entry is a definition, not a decision. Then ask "Shared understanding reached - start implementing?" Do not act before an explicit yes.
Versions
| Repository | Score | Licence |
|---|---|---|
| This page · BaSui01/snow-app-skills | 90 | MIT |
| nikospavlopoulos/devops-mutant-powers | 100 | MIT |
| RIJUDEVx/agent-skills | 100 | MIT |
| yuan88964/RelayStudio | 100 | MIT |
| SimplySylvia/Claude-Quorum | 100 | MIT |
| FrancoStino/opencode-skills-collection | 100 | MIT |
| cravingmaker/skills | 100 | MIT |
| helmholtzai-consultants-munich/helmholtz-agent-skills | 100 | Apache-2.0 |
| bendrucker/claude | 100 | MIT |
Database snapshot: 13.11.2026 · package version ddf8cc71e910 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100plan-grilling
Interrogate a plan, design or idea one round of questions at a time until every decision is settled. Use when stress-testing a plan before building, when a design feels underspecified, or turning a fuzzy workflow into a spec. - 86grill-me
Run a relentless interview of the user about a plan, design, or change — depth-first, one question at a time, each question carrying a recommended answer and code-checked before it is asked — until every branch of the decision tree is resolved. Use before any non-trivial change,… - 100interview
Interview the user in short rounds to turn a vague idea or plan into settled decisions, with a recommended answer for each question. Use before a spec, when requirements are fuzzy, or when the user says "grill me". - 84sw-grill-me
Interview the user relentlessly, in rounds, until every decision in a plan is settled; facts are your job, decisions are theirs. Use for "grill me", "interview me", "stress-test this plan". - 90interview
Grill the user about a plan, decision or idea, in rounds over a design tree, until nothing is left assumed. Use when the user wants to stress-test their thinking or says "grill me", and when another skill needs the user's decisions. - 100kit-grill
Ask the user the hard questions about an idea, plan or decision in their own terms, one decision per question, with a recommended answer first. Use in the explore skill's decide stage, for wayfinder grilling tickets, and whenever a plan has open decisions the user should make.…