@emseepea/testing 0.25.1 → 0.25.2

package/README.md

@@ -643,8 +643,9 @@ days.
 
 Import `@emseepea/testing/chatgpt` only when you want to test the real ChatGPT
 web experience. Ordinary framework tests do not enable this adapter or install
-Playwright. This is an experimental browser
-adapter: ChatGPT controls can change, and live host qualification is pending.
+Playwright. This is an experimental browser adapter: ChatGPT controls can
+change, so adopters must qualify each capability they rely on against their
+owned account and endpoint.
 
 On your own machine, install the optional browser tooling:
 
@@ -785,6 +786,16 @@ Return ordered, run-scoped calls matching the declared method, target, outcome,
 and optional result hash. Each call includes request/result SHA-256 values,
 endpoint SHA-256, sequence, timestamp, and `nativeClient`.
 
+Keep deterministic MCP qualification and native-host evidence separate.
+Command-line qualification can prove a resource link's shape, catalogue
+membership, exact `resources/read` result, and original bytes. A native ChatGPT
+journey separately proves whether ChatGPT materializes or rejects that link.
+
+Declare the expected `tools/call` and `resources/read` as distinct server calls.
+If ChatGPT does not issue the expected read, record the checkpoint as
+`incomplete` with category `server-observation-missing`; do not treat the
+tool-returned URI as proof of a native resource read.
+
 Set
 `nativeClient: "verified-chatgpt"` only when your authenticated server boundary
 establishes that the request came through the dedicated ChatGPT connection;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@emseepea/testing",
-  "version": "0.25.1",
+  "version": "0.25.2",
   "description": "MCP integration and semantic testing helpers",
   "license": "MIT",
   "repository": {