analyze-agent-codebase
Use to analyze an agent system's architecture or mechanisms from repository evidence, including follow-up questions and analysis…
Clarify a plan or design with one fixed Thread reviewer while maintaining confirmed domain language, current specifications and version decisions. Applies to the initiator and invited reviewer during that exchange; excludes solo questions, group debates and questioning without
$ npx -y skills add murray17/rovai-ai --skill grill-duo-with-docs --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/grill-duo-with-docsContext preview
The summary Claude sees to decide when to auto-load this skill.
Clarify a plan or design with one fixed Thread reviewer while maintaining confirmed domain language, current specifications and version decisions. Applies to the initiator and invited reviewer during that exchange; excludes solo questions, group debates and questioning without
name: grill-duo-with-docs description: Clarify a plan or design with one fixed Thread reviewer while maintaining confirmed domain language, current specifications and version decisions. Applies to the initiator and invited reviewer during that exchange; excludes solo questions, group debates and questioning without documentation work.
The initiator questions and maintains documents; one fixed partner independently advises without editing project documents. Investigate facts available from code, authoritative documents, tools, current input or Thread history. Ask the user for real choices. Use the user's language.
Use trusted sender identity, the triggering request and direct replies. A user start/answer or the current partner's direct reply to the valid invitation resumes the initiator. A direct request for this documentation variant makes you its reviewer only. Plain Grill Duo, old, invalid and late replies cannot advance, roll back or reopen this exchange.
Choose a relevant, available Thread partner other than yourself; address a trusted Agent ID and keep them throughout. Change only at the user's request, departure/unavailability, or a shift beyond their useful expertise; explain why. With none available, disclose solo questioning and keep the same round and documentation rules.
1. Prepare 1-4 independent questions with established prerequisites, numbered `Q1`-`Q4`; defer dependent questions. 2. Send the goal, confirmed facts, options, constraints and affected documents to the partner without your recommendation. 3. The partner returns one recommendation, main reason and risk per original number. They do not delegate, add questions or edit project files. 4. Present all open questions together, including tradeoffs, your recommendation, the partner's view, disagreements and affected documents. Ask the user to answer by number. 5. Maintain only confirmed content. Close questions when answered, cancelled or invalidated; start another round only after all close.
Keep unanswered questions, numbers and advice. Add no new questions mid-round. For a changed question, retain its number and re-review only that item; accept only a direct reply to the updated invitation. Partial answers confirm only the answered portion.
After dispatch, finish other current inputs and end while waiting. Follow CLI recovery on failure; do not blindly resend.
The initiator reads [Domain modeling](references/domain-modeling.md), [Glossary format](references/context-format.md) and [Decision routing](references/decision-routing.md) as needed. The partner need not load these authoring rules.
For each confirmed decision, determine whether to update domain vocabulary, Architecture, Contract, current-version decisions, implementation/acceptance notes, or no durable document. Apply repository rules. Unanswered, ambiguous or partner-only advice is not a confirmed fact. Do not create numbered ADR files.
When no important questions remain, summarize confirmed decisions, constraints, risks, document changes and nonblocking unknowns. Obtain the user's confirmation of the shared understanding before product implementation, then close the exchange. Unrelated messages and completed exchanges do not restart it.
Desktop and web workspace for lasting teams of coding agents, with shared conversations, task delegation, and collaborative memory.
Use to analyze an agent system's architecture or mechanisms from repository evidence, including follow-up questions and analysis…
Run a Thread discussion with several members to compare perspectives, options or tradeoffs and produce shared notes. Applies to…
Choose among Rovai messages, Missions, history and Memory; coordinate parallel member requests, multi-step operations or…
Disciplined diagnosis for hard, intermittent, or performance bugs. Use when the user explicitly asks to diagnose, debug, find…
Clarify or stress-test a plan, requirement, design or decision through user questions and one fixed Thread partner's independent…
Use to create a Rovai member or revise and confirm the unsaved identity card and avatar for that creation. Exclude profile…