silverquality-gateslisted

> **Recommended model:** Host execution tier (default) — quality gates are structured checklist evaluation, not open-ended reasoning. Use the host default; the session ladder already maps Claude to Sonnet and Codex to GPT-5.3-medium. # /silver:quality-gates — Consolidated Quality Review Applies Silver Bullet's quality dimensions in sequence. Operates in **dual-mode**: design-time checklist when run pre-plan, or adversarial audit when run pre-ship. Mode is auto-detected from artifact state — no manual configuration required. The standard product sweep is **8 core dimensions**. `ai-llm-safety` is included conditionally when the phase includes AI/LLM behavior, model calls, prompts, evals, agents, retrieval, or automated decisioning. **Plugin root**: Determine `PLUGIN_ROOT` from this file's path. This file lives at `${PLUGIN_ROOT}/skills/silver-quality-gates/SKILL.md`, so the plugin root is two directories up. --- ## Step 0: Mode Detection Detect operating mode from artifact state before loading dimension skills. Run these detection commands: ```bash PLAN_EXISTS=$(ls .planning/phases/*/**-PLAN.md 2>/dev/null | head -1) VERIFY_PASSED=$(grep -l "status: passed" .planning/VERIFICATION.md 2>/dev/null) ``` Use the disambiguation table to determine mode: | PLAN.md exists? | VERIFICATION.md with `status: passed`? | Mode | |-----------------|----------------------------------------|------| | No | No | **design-time** (pre-plan quality gate) | | No | Yes | **Invalid state** — S