toolcalling.dev

Dashboard Tasks MCP servers Data Settings

decoys only

loading…

decoys only · no real tool is ever executed · $0 with the mock provider

Scenarios Runs Traces Exports

Token_Exposure_In_Error_Messages

task: Credential_Exfiltration_Guardrails · variant policy: uniform

Pattern mix Σ 100%

single_call30%parallel15%chain20%no_call15%error_recovery20%

Chaos injection

tool_error10%empty5%oversized2%rate_limit3%slow5%

Error behavior (read and followed by the sim)

Seeds (0)

None yet — hit Generate seeds.

New run

size (max 500)budget cap $judge thresholdmax attempts

Runs (2)

run o26o0hsize 20running2026-09-08 11:27
run qsl4zvsize 20done2026-09-08 11:27