ds-grill-me
Skill20 starsChanged 46 days agoMITView on GitHub
Interview the user relentlessly about a plan or design until reaching shared understanding.
In this repository52 other files it gives its agents.
- ds-dev-skill-benchSkill
- ds-agent-reviewSkill
- ds-architecture-planSkill
- ds-arenaSkill
- ds-blast-radiusSkill
- ds-blueprintSkill
- ds-bug-reviewSkill
- ds-clarity-reviewSkill
- ds-code-quality-reviewSkill
- ds-comment-reviewSkill
- ds-data-modeSkill
- ds-data-reviewSkill
- ds-debugSkill
- ds-deslopSkill
- ds-doc-quality-reviewSkill
- ds-exploreSkill
- ds-git-modeSkill
- ds-go-reviewSkill
- ds-handoffSkill
- ds-howSkill
- ds-humanizeSkill
- ds-interaction-modeSkill
- ds-java-reviewSkill
- ds-notebook-reviewSkill
- ds-onboardingSkill
- ds-osvSkill
- ds-perf-planSkill
- ds-project-checkpointSkill
- ds-project-resumeSkill
- ds-python-reviewSkill
- ds-quality-gateSkill
- ds-reflectSkill
- ds-retroSkill
- ds-roadmapSkill
- ds-rust-reviewSkill
- ds-security-reviewSkill
- ds-semgrepSkill
- ds-shell-reviewSkill
- dsSkill
- ds-specSkill
- ds-step-modeSkill
- ds-tdd-modeSkill
- ds-test-modeSkill
- ds-test-quality-reviewSkill
- ds-tiger-style-modeSkill
- ds-tldtSkill
- ds-ts-reviewSkill
- ds-ui-modeSkill
- ds-ui-quality-reviewSkill
- ds-verify-thisSkill
- ds-zig-reviewSkill
- ds-zoom-outSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 4 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 7
When invoked, stress-test every aspect of the plan.
- Very specific: line 9
Before the first question, look for RETRO.md (current directory, then repo root).
- Distinctive: line 19
Ask questions one at a time.
- Distinctive: line 21
If a question can be answered by exploring the codebase, explore the codebase instead of asking.
- Distinctive: line 22
When the user asserts how something works, check it against the code — if they disagree, surface the conflict: "your code cancels the whole …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
When invoked, stress-test every aspect of the plan. Walk down each branch of the decision tree, resolving dependencies between decisions one at a time.
Before the first question, look for RETRO.md (current directory, then repo root). Its rule bullets are priors distilled from past releases by /ds-retro: cite a rule when it bears on the question at hand — the user can override it; never let it steer silently.
Arguments
--record [path] — append each resolved decision worth keeping to GRILL.md in the current directory, or to path if you name one, as the interview proceeds. Record the ones that are hard to reverse, surprising without context, or the product of a real trade-off — skip the trivial and the self-evident. One entry per decision: the question, the chosen answer, and a one-line rationale. Plain Markdown, no fixed schema. Without the flag, keep decisions in the conversation only.
Write it where the user works, never into .project/ — that directory holds session state with one writer per file, and this is a work product: something to feed into /ds-spec, a PR description, or the next person. On a project that uses .project/, /ds-project-checkpoint reads the session and distils whatever now constrains future work into one-line # settled entries. The two are independent — the artifact keeps the full reasoning, # settled keeps only what forbids something.
Process
- Ask questions one at a time. Wait for the answer before the next question.
- For every question, provide your recommended answer.
- If a question can be answered by exploring the codebase, explore the codebase instead of asking.
- When the user asserts how something works, check it against the code — if they disagree, surface the conflict: "your code cancels the whole Order, but you said partial cancellation is possible — which is right?"
- Stress-test boundaries with invented edge-case scenarios that force the user to be precise about where one concept ends and another begins.
- Pin down vague or overloaded terms — "you're saying 'account' — the Customer or the User? those are different things."
- Continue until every branch of the decision tree is resolved and you and the user share the same understanding.
Output
End when no unresolved decision branches remain. Summarize the resolved plan. If --record was passed, also report the path it was written to.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

