Help us improve
Share bugs, ideas, or general feedback.
From quorum
Triggers manual consensus audit process evaluating pending code change evidence from advocate/devil/judge roles, producing verdicts stored in SQLite.
npx claudepluginhub berrzebb/quorum --plugin quorumHow this agent operates — its isolation, permissions, and tool access model
Agent reference
quorum:agents/knowledge/protocols/auditThe summary Claude sees when deciding whether to delegate to this agent
Trigger the consensus audit process manually. Evaluates pending evidence and produces verdicts stored in SQLite. ``` quorum audit {{ arguments }} ``` Read config at `{ADAPTER_ROOT}/../../core/config.json`: - `audit_submit` MCP tool — evidence submission - `consensus.trigger_tag` / `agree_tag` / `pending_tag` — tag values - `consensus.roles` — provider-per-role mapping (advocate, devil, judge) |...Operates autonomous agent loops with clear stop conditions, progress tracking, and stall detection. Intervenes safely when loops stall or fail repeatedly.
Share bugs, ideas, or general feedback.
Trigger the consensus audit process manually. Evaluates pending evidence and produces verdicts stored in SQLite.
quorum audit {{ arguments }}
Read config at {ADAPTER_ROOT}/../../core/config.json:
audit_submit MCP tool — evidence submissionconsensus.trigger_tag / agree_tag / pending_tag — tag valuesconsensus.roles — provider-per-role mapping (advocate, devil, judge)| Flag | Effect |
|---|---|
--dry-run | Print the audit prompt without executing |
--no-resume | Start a fresh session (discard any saved session) |
--auto-fix | After audit, auto-correct rejected items |
--model <name> | Override auditor model (default from config) |
--reset-session | Delete saved session state before running |
--evidence <markdown> | Submit evidence directly (useful for worktree audits) |
Verdicts are stored in SQLite (not markdown files):
audit → provider reviews evidence
→ verdict stored via bridge.recordTransition()
→ audit-status.json marker written (fast-path for hooks)
→ quorum status shows result
Do NOT look for verdict.md or gpt.md — these files are eliminated. Use quorum status or quorum tool audit_history.
quorum tool audit_history --summary
| Context | Behavior |
|---|---|
| Interactive | Run audit, show verdict summary, suggest next steps |
| Headless | Run audit, output verdict, exit (caller reads audit-status.json) |
In headless mode, do NOT ask follow-up questions. Output the result and exit.
| Code | Meaning | Fix |
|---|---|---|
test-gap | Missing or insufficient tests | Add tests covering the claimed changes |
claim-drift | Evidence claim doesn't match actual diff | Update claim to match git diff |
scope-mismatch | Changed files not listed in evidence | Update Changed Files section |
quality-violation | Code quality check failed | Fix lint/type errors |