ADR Review Panel

What it does

This skill orchestrates six specialist reviewer subagents (context, technical, operational, security, cost, consequences) to audit an Architectural Decision Record (ADR). It normalizes input (Markdown or .docx), runs parallel reviewers, collects human-in-the-loop judgments, synthesizes findings, and emits a detailed PDF report plus a scannable PPTX deck. The output highlights top risks, tensions between reviewers, and concrete remediation suggestions.

When to use it

Invoke when you need structured, multi-perspective feedback on architecture proposals, design docs, RFCs, or ADRs—especially for decisions with production impact, compliance concerns, or long-term consequences. Use on drafts for formative feedback or on finalized ADRs for risk audits.

What's included

• Scripts: orchestration instructions and reviewer prompts live in references/ (references/reviewer_prompts.md, report_templates.md). The skill describes how to spawn six parallel subagents and collect structured findings.
• References: role prompts, report templates, and export conventions for PDF/PPTX.
• Instructions: prompt flow for clarifying questions, normalization of input formats, human-in-the-loop correction pass, synthesis heuristics, and output file layout.

Compatible agents

Designed for agent platforms that support spawning subagents and document generation (Claude Code, Cursor, Gemini Gems, and similar frameworks).