Skill

consensus-loop:guide

Guides writing evidence packages for consensus-loop watch file in code reviews. Covers claims, changed files, executable tests, results, risks, tag lifecycles, and rejection fixes.

ESLint

code-quality

Install

npx claudepluginhub berrzebb/claude-plugins --plugin consensus-loop

Tool Access

This skill uses the workspace's default tool permissions.

Preview

When submitting code changes for consensus review, write a properly structured evidence package in the watch file.

SKILL.md

Similar Skills

skill-lookup

Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.

prompts.chat

159.9k

prompt-lookup

Searches prompts.chat for AI prompt templates by keyword or category, retrieves by ID with variable handling, and improves prompts via AI. Use for discovering or enhancing prompts.

prompts.chat

159.9k

MCP Integration

6 files

Guides MCP server integration in Claude Code plugins via .mcp.json or plugin.json configs for stdio, SSE, HTTP types, enabling external services as tools.

plugin-dev

83.2k

Stats

Stars12

Forks3

Last CommitMar 18, 2026

Actions

View Source View Plugin View on GitHub View README

Consensus Loop — Evidence Package Guide

When submitting code changes for consensus review, write a properly structured evidence package in the watch file.

Step 0: Read Config

Read ${CLAUDE_PLUGIN_ROOT}/config.json first:

consensus.watch_file → submission path
consensus.trigger_tag / agree_tag / pending_tag → actual tag values
plugin.respond_file → auditor verdict file
plugin.locale → locale

Step 1: Write Evidence

Follow the format defined in ${CLAUDE_PLUGIN_ROOT}/templates/references/${locale}/evidence-format.md.

Required sections: Claim, Changed Files, Test Command, Test Result, Residual Risk.

Key rules:

Test Command must be executable as-is — no glob patterns, use explicit file paths
Test Result must be actual terminal output, not summaries
Claim must match the actual code changes
Changed Files paths must use backtick formatting
Every changed file must pass eslint individually
Use a single Write to the watch file (not sequential Edits)

Step 2: Tag Lifecycle

[trigger_tag] → auditor reviews → [agree_tag] or [pending_tag]
                                        ↓
                                Fix issues, re-submit with [trigger_tag]

Step 3: Addressing Rejections

When auditor returns [pending_tag]:

Read the rejection codes in the respond file (e.g., test-gap, claim-drift, scope-mismatch)
Fix each cited issue at the specific file:line locations
Update the evidence package with corrected claims, tests, and results
Keep the [trigger_tag] to trigger a new audit cycle

Full rejection code reference: ${CLAUDE_PLUGIN_ROOT}/templates/references/${locale}/rejection-codes.md