audit: deep platform audit, daily workflow, compliance flags, timestamped docs #3862
ccantynzcommented May 29, 2026
Originally written by @vercel[bot] on GitHub.
The latest updates on your projects. Learn more about Vercel for GitHub.
| Project | Deployment | Actions | Updated (UTC) |
|---|---|---|---|
| marco-reid-com | Preview, Comment | May 29, 2026 12:27am |
Cross-repo impact
See what breaks downstream if this PR merges.
⮌ Merged
This pull request was merged into main.
c comment · e edit title · m merge · a approve · r request changes · ? shortcuts
Originally written by @ccantynz-alt on GitHub.
Imported from https://github.com/ccantynz-alt/MarcoReid.com/pull/81
Summary
Deep audit of the entire platform — code vs. docs, what's built vs. what's needed, legal/compliance gaps, and competitive landscape research. Creates the infrastructure to keep docs in sync with code daily going forward.
What's in this PR
MASTER-AUDIT.md — Timestamped single source of truth for the platform. Code audit results, citation research (Stanford 2025 AI hallucination study), multi-jurisdiction compliance requirements, competitive landscape (Harvey/Clio/Smokeball/TR/Xero), and a ranked action plan. Every new container session reads this after CLAUDE.md so it never acts on stale information.
.github/workflows/platform-audit.yml— Daily automated audit at 6am NZST. Checks: Astra references, placeholder content, hardcoded secrets, brand casing violations, missing AI disclaimers, missing page metadata. Creates a GitHub issue if any check fails — new sessions have a clear action list waiting for them.Metadata fix —
/complianceand/legal-noticespages were the only marketing pages missing SEO metadata. Fixed vialayout.tsxwrappers (correct Next.js App Router pattern for"use client"pages).CLAUDE.md updates — 10 new timestamped decisions in Section 25. Stage Tracker updated (Stage 3 complete, Stage 4 in progress). Opening/Closing Ritual updated — MASTER-AUDIT.md is now step 2 of every session.
Key Findings From the Audit
Code is ahead of docs: The platform is ~85% production-ready. 84 API routes, 24 database models, Claude API integrated, Stripe fully wired, zero Astra references, zero lorem ipsum.
Immediate compliance flags:
Citation system: AI hallucination rates in best commercial tools: 17–43% (Stanford/Yale 2025). Required threshold before production: under 5%. Solution: CourtListener block-of-text API (free, 9M+ opinions) + Harvard CAP FAISS index (CC0, 6.7M opinions) as RAG corpus. Marco engine is built and ready — just needs API key + CourtListener wired in.
Most dangerous competitor: Smokeball in AU/NZ (4.8/5 stars, Archie AI, strong small-firm presence). Not Harvey — they won't touch the market we own.
Test plan
npm run build✅ confirmed clean/complianceand/legal-noticesworkflow_dispatchhttps://claude.ai/code/session_01BETeMQCCtyrema2QkKRWh6
Generated by Claude Code