Eval Criteria: Agent-EA v2 Engagement Playbook

Eval Criteria: Agent-EA v2 Engagement Playbook

Created: 2026-05-31 Target: production-lines/agent-ea/playbook/engagement.md Purpose: Operational runbook for running a lean, D1-first EA engagement end to end (stages, gates, contracts, conventions).

Scoring: each question is YES (1) or NO (0). Score = YES count / 12.

Questions

Clarity

  1. Is the golden rule (D1 is source of truth; corrections in data, never HTML) stated within the first 5 lines?
  2. Are the A-codes used in the Routing table either defined inline OR linked to their canonical source (Macroscope framework-guide)?

Structure

  1. Are the stages presented in execution order, with Routing at the head (Routing → Contract → Model → Publish → Review → Handover)?
  2. Does every stage have BOTH an explicit exit gate AND a named owner?

Operability

  1. For each automated stage (Model/Publish/Handover), is the exact run command (script + key flags) present OR explicitly cross-linked?
  2. Does the playbook define the first action for a brand-new request (the Routing scoping step)?
  3. Is the per-page data contract a concrete fillable checklist (not prose)?

Completeness

  1. Is the playbook free of [TODO]/placeholder/"to be filled" gaps?
  2. Does it cover follow-on mandates for an existing client (re-entry rule)?

Fidelity to the model

  1. Do all catalog/field conventions bind to the real D1 model (kind object/co, change_type on junction, epicId FK) without inventing fields?
  2. Does the Routing engagement→A-code bundle table match the canonical Macroscope framework-guide scenario bundles?

Traceability / risk

  1. Does the playbook surface known unresolved modeling issues (e.g. TES-CO vs editorial PROJ/CHG) and point to the spec/plan + engine README for deeper detail?