grill-duo
Skill255 starsChanged 16 days agoMITView on GitHub
Clarify or stress-test a plan, requirement, design or decision through user questions and one fixed Thread partner's independent review. Applies to the initiator and invited reviewer during that exchange; use grill-duo-with-docs when confirmed project documentation must also be maintained.
In this repository12 other files it gives its agents.
- AGENTS.mdAGENTS.md
- analyze-agent-codebaseSkill
- campfireSkill
- cli-operationsSkill
- diagnosing-bugsSkill
- grill-duo-with-docsSkill
- member-studioSkill
- memory-stewardshipSkill
- review-duoSkill
- tddSkill
- worktreeSkill
- writing-for-agentsSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Use trusted sender identity, the triggering request and direct replies:
- Distinctive: line 36
After dispatch, finish other current inputs and end while waiting for the reply.
- Distinctive: line 38
When no important questions remain, summarize the goal, decisions, constraints and major risks.
≈ “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
Grill Duo
The initiator asks questions; one fixed partner reviews independently. Investigate facts available in code, authoritative documents, tools, current input or Thread history. Ask the user for genuine choices. Use the user's language.
Roles and partner
Use trusted sender identity, the triggering request and direct replies:
- A user start/answer or the current partner's direct reply to the valid current invitation resumes the initiator.
- A direct Grill Duo review request makes you the partner for that request only.
- Old, invalid or late replies are supplementary; they cannot advance, roll back or reopen the exchange.
Choose a relevant partner who is not you, remains in the Thread and can receive work. Address a trusted Agent ID. Keep that partner unless the user requests a change, they leave or become unavailable, or the topic moves beyond their useful expertise; explain a change. With none available, disclose solo questioning and keep the same round rules.
One open round
- Prepare 1-4 independent questions with established prerequisites, numbered
Q1-Q4. Defer questions that depend on this round's answers. - Send the partner the goal, confirmed facts, options and constraints, without your recommendation.
- The partner returns one reply with a recommendation, main reason and risk for each original number. They do not delegate, add questions, decide for the user or implement.
- Present all open questions to the user together: choices, tradeoffs, your recommendation and the partner's view, including disagreements. Ask for answers by number.
- Close each question only when answered, cancelled or invalidated. Start the next round after all current questions close.
Keep unanswered questions, numbers and existing advice unchanged. Add no new questions mid-round. If the user changes a question, options or constraints, keep its number and re-review only that question; accept only a direct reply to the updated invitation. Partial answers close only the answered items.
Messages and completion
- Partner request:
rovai send --to <partner-agent-id> --body <questions>. - Partner response:
rovai send --to <requester-agent-id> --body <advice>. - User questions or final confirmation:
rovai send --public-only --to-user --body <questions-or-summary>.
After dispatch, finish other current inputs and end while waiting for the reply. Follow CLI recovery on failure; do not blindly resend.
When no important questions remain, summarize the goal, decisions, constraints and major risks. Obtain the user's confirmation of that shared understanding before implementation, then close the questioning exchange. Solo questions, group debates, unrelated messages and closed exchanges do not start this workflow.
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.

