@miraigent/free-ai-ops-mcp 0.1.31 → 0.1.33

package/CHANGELOG.md

@@ -1,5 +1,21 @@
 # Changelog
 
+## [0.1.33] - 2026-07-08
+
+### Changed
+
+- Added a minimal README parsing example that shows MCP users how to read the
+  JSON tool result from `content[0].text` before wiring the package into a
+  desktop client or private workflow.
+
+## [0.1.32] - 2026-07-07
+
+### Added
+
+- Added smoke-test assertions that parse MCP tool results from
+  `content[0].text` as JSON and verify exact public fields for the human review
+  and prompt risk JSON-RPC sessions.
+
 ## [0.1.31] - 2026-07-06
 
 ### Changed

package/README.md

@@ -136,6 +136,26 @@ files.
 After the smoke test runs, look for the status field from the tool you called
 before changing client configuration or private workflows.
 
+In raw JSON-RPC output, MCP wraps each tool result inside `content[0].text`.
+Parse that text value as JSON before you check the returned decision:
+
+    const response = JSON.parse(lineFromStdout);
+    const toolResult = JSON.parse(response.result.content[0].text);
+    console.log(toolResult.gateStatus);
+
+For the bundled smoke-test session, that prints `stop`. Use the parsed
+`gateStatus`, `recommendation`, or `recommendedStatus` field from the relevant
+tool before changing a desktop MCP client config or connecting private data.
+
+If you only need a terminal check, you can copy the final JSON-RPC line from the
+smoke-test output and inspect the wrapped text field with Node:
+
+    node -e 'const r=JSON.parse(process.argv[1]); console.log(JSON.parse(r.result.content[0].text));' '<paste one response line here>'
+
+Paste only the synthetic response from the public example. Do not paste private
+customer records or desktop client logs into shell history, screenshots, or
+GitHub issues.
+
 For `human_review_gate`:
 
 - `auto_ok`: the synthetic draft looks low risk, but the tool still does not

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@miraigent/free-ai-ops-mcp",
-  "version": "0.1.31",
+  "version": "0.1.33",
   "description": "Free MCP server for developers adding human review gates, prompt risk checks, FAQ review, and CRM note safety to AI tools.",
   "homepage": "https://github.com/Miraigent/miraigent-free-ai-ops-mcp",
   "type": "module",

package/scripts/smoke-test.mjs

@@ -174,6 +174,11 @@ assert.deepEqual(
 assert.match(sampleResponses[2].result.content[0].text, /gateStatus/);
 assert.match(sampleResponses[2].result.content[0].text, /reviewOwner/);
 assert.match(sampleResponses[2].result.content[0].text, /boundary/);
+const sampleToolResult = JSON.parse(sampleResponses[2].result.content[0].text);
+assert.equal(sampleToolResult.tool, 'human_review_gate');
+assert.equal(sampleToolResult.gateStatus, 'stop');
+assert.equal(sampleToolResult.nextLogRow.gate_status, 'stop');
+assert.equal(sampleToolResult.boundary, 'This tool is a review helper. It does not send messages.');
 
 const promptRiskSession = await readFile(
   new URL('../examples/prompt-risk-json-rpc-session/sample-session.jsonl', import.meta.url),
@@ -192,3 +197,11 @@ assert.match(promptRiskResponses[2].result.content[0].text, /prompt_risk_review/
 assert.match(promptRiskResponses[2].result.content[0].text, /stop_before_ai_use/);
 assert.match(promptRiskResponses[2].result.content[0].text, /customer_facing/);
 assert.match(promptRiskResponses[2].result.content[0].text, /sensitive_data_possible/);
+const promptRiskToolResult = JSON.parse(promptRiskResponses[2].result.content[0].text);
+assert.equal(promptRiskToolResult.tool, 'prompt_risk_review');
+assert.equal(promptRiskToolResult.recommendation, 'stop_before_ai_use');
+assert.deepEqual(promptRiskToolResult.riskFlags, ['customer_facing', 'sensitive_data_possible']);
+assert.equal(
+  promptRiskToolResult.boundary,
+  'This tool is a prompt risk helper. It is not legal advice and does not call an AI API.'
+);