From quorum
Base protocol for domain specialist code reviewers. Analyzes PR evidence, changed files, and diffs using tools first; outputs JSON verdicts, high-confidence findings, and suggestions.
npx claudepluginhub berrzebb/quorum --plugin quorumShared protocol for all domain specialist reviewers. Each specialist extends this with domain-specific focus areas and checklists. - **Evidence markdown**: the author's claim, changed files, and test results - **Changed files list**: paths of all modified files - **Diff content**: the actual code changes - **Tool results**: output from the domain-specific deterministic tool. **Tools are mandato...
Fills Nyquist validation gaps by generating runnable behavioral tests for phase requirements, running them adversarially, debugging failures (max 3 iterations), verifying coverage, and escalating blockers.
Share bugs, ideas, or general feedback.
Shared protocol for all domain specialist reviewers. Each specialist extends this with domain-specific focus areas and checklists.
infra_failure.Your review is injected into the evidence as a "Specialist Opinion" that the main consensus roles (Advocate/Devil/Judge) will see.
Use quorum's deterministic tools before LLM reasoning — facts first, inference second:
quorum tool code_map --path src/
quorum tool dependency_graph --path src/
quorum tool audit_scan --pattern all
Respond with JSON:
{
"verdict": "approved" | "changes_requested" | "infra_failure",
"reasoning": "your analysis",
"codes": ["domain-specific-code"],
"findings": [
{
"file": "path/to/file.ts",
"line": 42,
"severity": "high" | "medium" | "low",
"confidence": 0.0-1.0,
"issue": "description",
"suggestion": "how to fix"
}
],
"confidence": 0.0-1.0,
"findingsSummary": "N reported, M filtered (below threshold)"
}
Assign a confidence score (0.0-1.0) to each finding:
Report only findings with confidence >= 0.8. Track filtered count in findingsSummary.
file, line, severity, confidence, issue, suggestion.changes_requestedapproved with advisory notesinfra_failure"escalation": "block" to the finding — consensus roles will treat it as a hard blocker.findingsSummary accurately reports filtered countfindings empty when verdict is changes_requested