docs: define enterprise-readiness test harness boundary

Author: LalaSkye

enterprise-execution-readiness/docs/enterprise-readiness-test-harness.md

@@ -0,0 +1,111 @@
+# Enterprise-Readiness Test Harness
+
+## Status
+
+**Version:** v0.1  
+**Scope:** Synthetic / mocked downstream systems / CI replay  
+**Claim:** Test-harness evidence only  
+
+This document defines what the enterprise-shaped scenario harness currently proves and what remains absent.
+
+## Clean claim
+
+This harness tests whether a bounded AI-assisted action can reach a mocked downstream system when required authority is missing.
+
+For ESP-001, the answer is:
+
+```text
+missing authority_token
+  -> DENY
+  -> mock email adapter not called
+  -> downstream_send=false
+  -> receipt_written=true
+```
+
+## What this harness proves
+
+On the synthetic path, the harness proves:
+
+- the action class is named: `SEND_EXTERNAL_EMAIL`
+- the required authority field is missing: `authority_token`
+- the gate returns `DENY`
+- the mocked downstream email adapter is not called
+- the adapter records zero sent messages
+- a receipt is generated
+- the scenario can be replayed by pytest
+- the test can run in GitHub Actions
+
+## What this harness does not prove
+
+This harness does not prove:
+
+- live SMTP / API non-execution
+- production runtime enforcement
+- enterprise deployment
+- certification
+- compliance
+- legal adequacy
+- closure of all bypass paths
+- no dispatch through queues, retries, alternate APIs, human handoff, or connector side channels
+- real-world controlled application
+- external review
+
+## Required evidence for stronger claims
+
+| Stronger claim | Required evidence |
+|---|---|
+| Live runtime enforcement | Gate integrated with a real execution layer or realistic service boundary |
+| Downstream non-execution | External connector call logs or service-level event evidence |
+| Path-universal control | Architecture showing every consequence path routes through the gate or is explicitly out of scope |
+| Enterprise readiness | Multiple realistic enterprise workflows, CI replay, bypass tests, external review, and controlled organisational pilot evidence |
+| Production readiness | Deployment architecture, operational monitoring, failure handling, persistence, security review, and live audit trail |
+| Certification / compliance | Independent qualified assessment against a named framework or legal requirement |
+
+## Current label
+
+Safe:
+
+```text
+enterprise-readiness test harness
+```
+
+Unsafe:
+
+```text
+enterprise-ready system
+enterprise deployment evidence
+production enforcement proof
+certified control
+compliance-ready system
+```
+
+## Next hardening path
+
+1. Add more mocked consequence classes:
+   - payment instruction
+   - access change
+   - HR decision support
+   - vendor risk escalation
+
+2. Add bypass tests:
+   - retry cannot dispatch after denial
+   - queue cannot dispatch after denial
+   - stale DecisionRecord rejected
+   - alternate send path blocked
+   - audit failure fails closed
+
+3. Add evidence matrix:
+   - claim
+   - required evidence
+   - current proof
+   - missing proof
+   - next test
+
+4. Add external review route:
+   - issue template for hostile review
+   - reviewer checklist
+   - reproducibility instructions
+
+## Stop rule
+
+Do not call this enterprise-ready until an external reviewer or bounded organisation can inspect, run, and challenge the harness against a realistic workflow.