sample report

One artifact for engineering, security, and release.

This is what a team gets back: client behavior, raw trace, security surface, verdict, and a receipt that can be shared outside the product.

run_4821 not fit to ship

cursor streaming failure / high path finding / sealed evidence

report artifact

filesystem-mcp@1.4.0

run #4821 / 47s / two live clients / security review

NOT FIT TO SHIP1 client fail / 1 high finding

clientconnectdiscoverschematoolserrorsstreamresourcesparallelreconnect claude codePASSPASSPASSPASSPASSPASSPASSPASSPASS cursorPASSPASSPASSPASSPASSFAILPASSPASSPASS

{
  "client": "cursor",
  "step": "streaming",
  "request": {"jsonrpc": "2.0", "method": "tools/call"},
  "response": {"error": {"code": -32001, "message": "request timed out after 5000ms"}},
  "divergence": "claude-code completed same call in 1188ms"
}

replay:
npx throne replay run_4821 --client cursor --step streaming

expected:
stream close event before 5000ms

observed:
timeout without surfaced tool error

THR-PATH-01HIGH

schema.properties.path accepts relative segments without root normalization.

THR-PROMPT-07MEDIUM

tool output enters model context without delimiter guard.

THR-RATE-09LOW

write tool has no request-rate guard in schema metadata.

scan idrun_4821

targetgithub.com/acme/files-mcp

sealed at2026-06-13 08:42 UTC

evidence hashsha256:09bed78a953f...

why buyers care

A report turns MCP risk into a release decision.

eng

Fix the exact failure

The failing cell links to the trace, the replay command, and the observed client behavior.

sec

Review the exact surface

Security findings show severity, location, reproduction, and fix guidance.

exec

Decide without debate