CASE · 01
A user has drafted their endorsement application and wants conservative feedback before submission
Grades a draft UK Global Talent Visa application against the four endorsement criteria — Recognition, Innovation, Impact, Contribution — and returns an A+ to F grade with per-criterion rationale.
“I can grade your draft UK Global Talent Visa application against the four endorsement criteria and tell you which evidence is weakest.”
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "score_draft",
"arguments": {
"form_data": {
"fullName": "Priya R.",
"cvText": "Staff ML engineer, 9y. Led ranking team of 12…",
"awards": "Best Paper, NeurIPS 2023",
"publications": "First author, 4 papers (ICML, NeurIPS)",
"references": "CTO, UK fintech; Professor, Imperial College"
}
}
}
}{
"overallGrade": "B",
"overallScore": 74,
"criteria": {
"recognition": {
"grade": "B",
"note": "One major award; add press or invited talks."
},
"innovation": {
"grade": "A",
"note": "Strong first-author publication record."
},
"impact": {
"grade": "C",
"note": "Quantify downstream product or user impact."
},
"contribution": {
"grade": "B",
"note": "Reference letters cover two of three criteria."
}
},
"weakest": "impact"
}Rate limit. score_draft is backed by a shared free-tier model quota, so it is capped at 30 calls per IP per 24 hours. When the cap trips, the JSON-RPC error includes data.rate_limited=true.