paperclip/evals/promptfoo/mcp-gateway-run-summary.md

1.7 KiB

MCP Gateway Promptfoo Run Summary

Date: 2026-06-16 Issue: PAP-11188

Suite

  • Config: evals/promptfoo/promptfooconfig.yaml
  • Cases: evals/promptfoo/tests/mcp-gateway.yaml
  • Provider used for local baseline: echo override
  • Filter: ^mcp_gateway\.

Command

npx promptfoo@0.103.3 eval -c evals/promptfoo/promptfooconfig.yaml --providers echo --filter-pattern '^mcp_gateway\.' --no-cache --no-progress-bar --no-write -o evals/promptfoo/mcp-gateway-results.json

Config validation:

npx promptfoo@latest validate -c evals/promptfoo/promptfooconfig.yaml

Result

  • Total cases: 12
  • Passed: 12
  • Failed: 0
  • Errors: 0

Coverage

  • allow / profile_allows_tool read-only success behavior
  • 403 deny_default denied unsafe tool behavior
  • 409 approval_required pending gateway approval behavior
  • 409 action_not_approved rejected approval behavior
  • 409 formal_approval_required formal board approval wait behavior
  • 429 rate_limited rate-limit handling behavior
  • 422 remote_http_missing_secret credential repair without secret leakage
  • 401 session_revoked stale gateway-token handling
  • remote HTTP header forwarding and redacted credential audit evidence
  • exact named/on-demand gateway target selection
  • elicitation-required human input wait path
  • 409 approved_tool_target_changed stale approval / target drift behavior

Residual Risk

No provider API keys were present in the heartbeat environment, so the live OpenRouter model matrix was not executed. The local baseline proves promptfoo wiring and deterministic assertions. Live model scoring should be run with OPENROUTER_API_KEY or equivalent before using this as a release gate.