From 4ace3012e010a64f4345673434e41355465d9ff1 Mon Sep 17 00:00:00 2001 From: Cursor Agent Date: Fri, 25 Sep 2026 01:50:39 +0000 Subject: [PATCH] Show the GO/NO-GO banner in the report overview The opening diagram now matches the HTML report: readiness evidence comes first, and the run stays green unless enforcement is requested. Co-authored-by: Dor Ingber --- README.md | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index ebd1263..783696e 100644 --- a/README.md +++ b/README.md @@ -34,12 +34,15 @@ ## The report is the product -Most agent tests dump JSON into CI logs nobody reads. **agent-test-kit** generates a single self-contained HTML file — tool-flow strips, MCP chips, prompt review diffs, golden guard outcomes — the same UI whether you ran mock tests or live E2E. +Most agent tests dump JSON into CI logs nobody reads. **agent-test-kit** generates a single self-contained HTML file — a GO / NO-GO readiness banner, tool-flow strips, MCP chips, prompt review diffs, golden guard outcomes — the same UI whether you ran mock tests or live E2E. ``` ┌─ Agent test report ─────────────────────────────────────────────┐ │ Agent: billing-agent Env: integration │ ├─────────────────────────────────────────────────────────────────┤ +│ Release readiness NO-GO │ +│ Blocking: persona "novice" pass rate 60% is below 80% │ +├─────────────────────────────────────────────────────────────────┤ │ Scenarios 6 Passed 6 Tool calls 14 Tokens 2.1k │ ├─────────────────────────────────────────────────────────────────┤ │ MCP tool flow │ @@ -49,10 +52,12 @@ Most agent tests dump JSON into CI logs nobody reads. **agent-test-kit** generat │ │ get_inv… │ │ get_iss… │ │ create… │ │ │ └──────────┘ └──────────┘ └──────────┘ │ ├─────────────────────────────────────────────────────────────────┤ -│ ▾ Assertions ▾ Tool calls ▾ Timeline ▾ Side effects │ +│ ▾ Assertions ▾ Tool calls ▾ Timeline ▾ Side effects │ └─────────────────────────────────────────────────────────────────┘ ``` +The banner is the release evidence: failed scenarios, security failures, duplicate writes, and any persona or baseline threshold you set. The pytest run stays green unless you pass `--agent-enforce-readiness`. + ```bash pytest tests/ -v \ --agent-report-json=reports/results.json \