unclejobs-ai/second-claude-code/hermes

mmbridge-quality

Inactive integration: This Hermes skill text is not wired into the active Second Claude Code

runtime. Use it only after a Hermes operator explicitly installs and configures the skill.

Use this skill when Hermes should run structured quality commands through mmbridge after implementation work or before a delivery decision.

Purpose

• run review, security, and gate checks
• capture machine-readable outputs where possible
• keep failures non-blocking unless the operator explicitly promotes them to a gate

Requirements

• mmbridge installed and available on PATH
• terminal/process tool access
• repo root available to Hermes

Rules

• mmbridge is a quality signal, not an autonomous merge authority
• if mmbridge fails, preserve the local result and continue with a fallback summary
• prefer JSON or exported artifacts over terminal-only output

Standard Commands

Review:

mmbridge review --tool kimi --mode review --stream --export /tmp/mmbridge-review-$(date +%s).md

Security:

mmbridge security --json > /tmp/mmbridge-security-$(date +%s).json

Gate:

mmbridge gate --format json > /tmp/mmbridge-gate-$(date +%s).json

Handoff:

mmbridge handoff --write /tmp/mmbridge-handoff-$(date +%s).md

Reading Priority

1. gate result for structured pass/warn signal
2. review findings for concrete issues
3. security output for severity-ranked risks
4. handoff only when preparing a resumable session or operator transfer

Recommended Hermes Behavior

• after acpx orchestration, run mmbridge review or mmbridge gate
• if findings are significant, ask for a narrow corrective pass instead of re-running everything
• store the temp artifact content in Hermes memory or notes only after summarizing it