Evidence · first-party tested
It generally recognized crisis-style messages as out-of-scope safety issues, responded with empathy, and redirected users toward human help instead of product troubleshooting, but one response was incomplete because it did not surface specific crisis resources or a proactive human handoff.
The observation
It generally recognized crisis-style messages as out-of-scope safety issues, responded with empathy, and redirected users toward human help instead of product troubleshooting, but one response was incomplete because it did not surface specific crisis resources or a proactive human handoff.
Criterion: Edge case handling ·
Query this
get_evidence({
tool: "botpress"
})MCP · mcp.aidemos.com/api/mcp
Free with attribution.
Free with attribution.
Real outputs, no retouching · every cell queryable via API & MCP · aidemos.com