docs/snapshots/2026-08-01-v0.1.5.md

# 2026-08-01 — version 0.1.5 release-evidence snapshot

- Plugin version: `0.1.5`
- Source branch: `agent/endpoint-security-benchmarks`
- Snapshot class: `local_validation`
- Validation evidence: 235 offline regression tests, 311 endpoint fuzz cases,
  8 security-audit checks, and 10.0/10 deterministic readiness
- Release archive: `empire-llm-codex-0.1.5.zip`
- Sensitive-data review: no credentials, organization identifiers, private
  prompts, raw provider responses, or billing payloads included

## Shipped boundary

Version `0.1.5` adds three-state credential reporting, verified keyring
readback, paid-response recovery, and measurement-only context preflight
receipts. Context recommendations do not alter provider dispatch, and paid
partial results never trigger an automatic retry or continuation.

## Distribution boundary

OpenAI App Store approval is recorded by publisher attestation for the
submission lineage. This snapshot is not a live portal capture and does not
prove public directory visibility, regional availability, search placement, or
per-account access.

## Visual evidence boundary

The marketplace and bounded-review images referenced by the repository remain
historical `0.1.2` product illustrations. They are retained to demonstrate the
intended interaction contract, not as representations of every `0.1.5` change.