Stats
Actions
Tags
Help us improve
Share bugs, ideas, or general feedback.
From grill-each-other
Code-aware variant of grill-me-agents — interrogates a multi-agent design while continuously cross-referencing existing agent definitions, skills, prompts, and tool configs in the repo. Use ONLY when the agent stack is already implemented in the repo and you want to stress-test a *change* to it (versus designing from scratch — use `grill-me-agents` for greenfield). Triggers: "grill the existing agent stack", "stress-test the agent stack", "grill the implemented agents", "audit the agent topology against code". Do NOT use for greenfield agent design (use `grill-me-agents`) or for two agents grilling each other (use `peer-grill`).
npx claudepluginhub jackreis/arbiter --plugin grill-each-otherHow this skill is triggered — by the user, by Claude, or both
Slash command
/grill-each-other:grill-me-with-agentsThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Interview me relentlessly about every aspect of the proposed agent collaboration until we reach a shared understanding. Walk down each branch of the design tree, resolving dependencies one-by-one. For each question, provide your recommended answer.
Provides PHI/PII compliance patterns for healthcare apps including data classification, row-level security access control, audit trails, encryption, and common leak vectors. Useful for patient data features, APIs, and code reviews.
Share bugs, ideas, or general feedback.
Interview me relentlessly about every aspect of the proposed agent collaboration until we reach a shared understanding. Walk down each branch of the design tree, resolving dependencies one-by-one. For each question, provide your recommended answer.
Ask the questions one at a time.
Before asking any branch, scan the repo for ground truth and cite what you found. Look in (in priority order):
.claude/agents/, .claude/shared-agents/ — existing agent definitions.claude/skills/, .claude/shared-skills/ — skills the agents may invoke.claude/settings.json, .claude/settings.local.json — permissions, hooks, env.claude/AGENT-STACK.md, CLAUDE.md, any AGENT_DESIGN.md — declared topology.github/workflows/ — CI hooks that gate or trigger agent runsIf a branch can be answered from those sources, propose the answer for my confirmation rather than asking from scratch. Quote the file and line so I can verify.
If the codebase contradicts what I tell you, surface the contradiction and ask which is canonical before proceeding.
Cover the same 13 branches as grill-me-agents, in order, only descending once an upstream branch is settled:
tools: allowlists in agent frontmatterAskUserQuestion, ExitPlanMode, and confirmation gatesmodel: frontmatter)Do not move to the next branch until the current one has a concrete answer, not a placeholder. If I give a vague answer, push back with the specific file or edge case that makes it vague.
When every branch is resolved, summarize the agreed design and offer to write it to AGENT_DESIGN.md using the template at ~/.claude/skills/grill-me-agents/AGENT_DESIGN.template.md (sister-skill template, intentionally shared). Include: roster (with file paths), topology diagram (ASCII), handoff contracts, failure/termination rules, observability hooks, and a diff against the previous design if one existed.
grill-me-agents — same 13 branches but greenfield (codebase-blind). Use when designing a new agent stack from scratch.grill-me — single-track plan/design grilling (not agent-specific).peer-grill — for two or more agents to grill each other (file-based reconciliation protocol).