brutal-grill
Skill2 starsChanged 4 months agoMITView on GitHub
Repo-grounded requirements interrogation that resolves related planning decisions in batched rounds. Use when a feature, refactor, or system change still has material product or engineering choices, or when another Brutal planning skill needs a decision-complete requirements brief.
In this repository25 other files it gives its agents.
- brutal-deepresearchSkill
- brutal-idea-evalSkill
- brutal-inf-fix-loopSkill
- brutal-planSkill
- brutal-pr-finding-fixerSkill
- brutal-pr-fix-loopSkill
- brutal-project-reviewSkill
- brutal-pr-reviewSkill
- brutal-realitySkill
- brutal-reviewSkill
- brutal-specSkill
- brutal-swarmSkill
- brutal-ticketsSkill
- brutal-wayfinderSkill
- brutal-workerSkill
- codex-coopSkill
- design-jepa-modelsSkill
- factory-designSkill
- funcvec-dedupe-fixerSkill
- grill-meSkill
- handoffSkill
- opencode-delegateSkill
- rust-architecture-reviewSkill
- rust-pr-reviewSkill
- supply-chain-analyzerSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 2 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 13
Read applicable AGENTS.md, CLAUDE.md, TARGET.md, and referenced workflow documents.
- Very specific: line 15
Read CONTEXT.md or the relevant entry from CONTEXT-MAP.md, plus ADRs in the affected area.
- Distinctive: line 24
Ask every currently known, relevant, independent question in one round.
- Distinctive: line 30
Use structured user input when it is available, the whole round has at most three questions, and each question has 2-3 mutually exclusive …
- Very specific: line 70
Do not persist planning artifacts or edit glossary and ADR files.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
Brutal Grill
Resolve every material planning decision without asking the user for facts the repository can supply.
Ground The Discussion
- Read applicable
AGENTS.md,CLAUDE.md,TARGET.md, and referenced workflow documents. - Read
CONTEXT.mdor the relevant entry fromCONTEXT-MAP.md, plus ADRs in the affected area. Use their domain language and surface contradictions. - Inspect affected modules, callers, schemas, migrations, interfaces, tests,
and prior art with
rgandrg --files. - Separate discoverable facts from decisions. Investigate facts; ask only for decisions.
Grill In Decision Rounds
Ask every currently known, relevant, independent question in one round. Two questions are independent only when neither answer can change the other's necessity, wording, or options. After each response, investigate newly exposed branches and batch the questions they make necessary. Ask nothing when repository evidence resolves every decision.
Use structured user input when it is available, the whole round has at most
three questions, and each question has 2-3 mutually exclusive options. If the
tool is unavailable or the round does not fit, use one numbered questionnaire
and request a compact reply such as 1B 2A 3: custom answer.
For each decision question:
- give only the minimum repository evidence that makes it necessary
- offer meaningful options with one-line tradeoffs
- recommend one option in one line
Give a default or example, not an invented recommendation, for genuinely open-ended input. Keep planning questions separate from caller-owned final approval, write permission, and safety or model-fallback consent.
Resolve contradictions immediately. Prefer concrete scenarios, payloads, state transitions, failure examples, and boundary values over abstract questions. Continue until the brief covers:
- goal, actors, and observable behavior
- inputs, outputs, interfaces, and compatibility constraints
- scope, anti-goals, and existing behavior that must remain stable
- empty, invalid, partial-failure, concurrency, and dependency-failure behavior
- security, privacy, performance, observability, migration, and rollout where relevant
- acceptance scenarios and 1-3 invariants
Skip categories that have no bearing on the work. Never turn the list into a generic questionnaire. Resolve low-impact choices from repository precedent and record the assumption; ask only when the choice materially changes behavior, scope, risk, compatibility, or delivery.
Return A Requirements Brief
Return a compact brief of confirmed requirements, boundaries, invariants, constraints, assumptions, acceptance scenarios, and unresolved decisions. Reference evidence instead of repeating the investigation or conversation. Mark the brief decision-complete only when every high-impact branch is resolved and the repository evidence agrees with it.
Do not persist planning artifacts or edit glossary and ADR files. Pass proposed
domain-language or architectural changes to brutal-spec for approval.
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.

