---
title: Appendix A — Blind Comparison Workflow Schema Additions
type: claude-md-appendix
parent: CLAUDE.md
created: 2026-05-08
status: proposal
---

# Appendix A — Schema additions proposed for §4.5 Blind Comparison Workflow

**These are PROPOSALS only. Do not add to Airtable without Rocky-confirmed schema-change task.**

## A.1 New table: `Blind Comparison Sessions`

Scope: operational (per-comparison, per-job). Lives in LFCS-Operations base.

| Field | Type | Required | Description |
|---|---|---|---|
| `session_id` | singleLineText (primary, UUID) | Yes | |
| `job` | multipleRecordLinks → Jobs | Yes | |
| `bid_a` | multipleRecordLinks → Bids | Yes | The comparator's prior bid |
| `bid_b` | multipleRecordLinks → Bids | Yes | The new blind-priced bid |
| `comparator_name` | singleLineText | Yes | E.g. "Hasibul Khan", "Liam Fitzgerald" |
| `new_pricer_name` | singleLineText | Yes | E.g. "claude-code-opsman + Rocky" |
| `comparator_artefact_paths` | multilineText | Yes | One per line; agent-excluded paths |
| `quarantined_methodology_record_ids` | multilineText | No | One per line; methodology rows excluded for this session |
| `blind_phase_started_at` | dateTime | Yes | |
| `blind_phase_ended_at` | dateTime | No | Reveal-phase trigger |
| `comparison_capture_at` | dateTime | No | When comparison MD written |
| `unquarantined_at` | dateTime | No | When methodology rows returned to active pool |
| `pre_comparison_sealed_reasoning` | multilineText | Yes | Sealed agent reasoning trail |
| `post_comparison_corrections_text` | multilineText | No | Populated step 8 |
| `derived_methodology_corrections` | multipleRecordLinks → Bid Line Item Overrides | No | Override rows derived from corrections |
| `status` | singleSelect | Yes | Active / Sealed / Reveal / Corrections-Pending / Closed / ABORTED-BREACH |

## A.2 New field on `_methodology_Rate_Templates` + `_methodology_Productivity_Benchmarks`

| Field | Type | Required | Description |
|---|---|---|---|
| `_internal_quarantined_for_session_id` | singleLineText | No | Set to blind-comparison session UUID during quarantine; cleared on unquarantine |

(Confidential per `_internal_*` prefix rule; never sanitised out, never rendered to external.)

## A.3 New field on `Bid Line Item Overrides`

| Field | Type | Required | Description |
|---|---|---|---|
| `linked_blind_comparison_session` | multipleRecordLinks → Blind Comparison Sessions | No | Set when override emerged from a blind-comparison reveal phase |

**Rocky decision needed before any of A.1–A.3 land in Airtable.** Until then, blind comparison runs partially: workflow §4.5 steps 1–6 + 9 work via vault md only (`_Internal/blind-comparison-{job}-{date}.md`); steps 7–8 need the Sessions table to capture structured corrections.