Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Prioritize inbound client calls by reading transcripts, classifying urgency, matching callers to legal client and matter context, and preparing a callback list for human review.
.claude/skills/zapier-mcp-urgent-caller-triage/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 124% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 168% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 80% | 0% |
Use this skill when a law firm needs to identify which inbound callers need attention first and give the callback owner context before they call back.
| Tier | Signals | Response target | | --- | --- | --- | | Critical | Court date or filing deadline within 48 hours, opposing counsel deadline, safety concern, client threatening to leave | Immediate review | | Urgent | Billing dispute, repeated unreturned calls, frustrated language, time-sensitive filing, escalation request | Same-day review | | Standard | Status check, document request, scheduling, general question | Next-business-day review |
Tune the signal list by practice area. Family law, personal injury, criminal defense, immigration, estate planning, and business law teams will each need different examples and thresholds.
For each flagged call, return the minimum useful callback entry by default:
text[Tier] - <Masked Caller Identifier> Call ID: <Call ID or secure source link> Reason: <Neutral reason this call was flagged> Deadline category: <Immediate, same day, next business day, or none found> Owner: <Assigned attorney or staff owner> Recommended next step: <Callback, manual lookup, escalation, or standard follow-up>
Include full phone numbers, matter names, matter status, billing details, transcript excerpts, or deadline specifics only inside approved legal systems or access-controlled firm channels where that level of detail is explicitly required. Use billing context only as a coarse flag such as "billing review requested"; do not include invoice amounts, payment details, or dispute narratives in broad callback summaries.
Critical calls should appear first. Within each tier, sort by deadline proximity, then by repeat-caller count.
This skill is generalized from Iron Noodle's legal automation work using Zapier MCP. Read the full story: How an automation consultancy uses Zapier to rewire law firms in 9 days.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 11,461 | 11,922 | +4% | 1 | 1 | 0% | 1,892 | 2,848 | +51% | 0 | 0 | — |
case-02 | pass→pass | 9,606 | 11,145 | +16% | 1 | 1 | 0% | 1,564 | 2,621 | +68% | 0 | 0 | — |
case-03 | pass→fail | 10,974 | 5,226 | -52% | 1 | 1 | 0% | 1,820 | 1,664 | -9% | 0 | 0 | — |
case-04 | fail→pass | 7,564 | 3,540 | -53% | 1 | 1 | 0% | 1,285 | 1,423 | +11% | 0 | 0 | — |
case-05 | fail→pass | 4,765 | 4,810 | +1% | 1 | 1 | 0% | 749 | 1,675 | +124% | 0 | 0 | — |
case-06 | fail→pass | 3,903 | 4,416 | +13% | 1 | 1 | 0% | 584 | 1,566 | +168% | 0 | 0 | — |
case-07 | pass→pass | 3,739 | 4,155 | +11% | 1 | 1 | 0% | 556 | 1,609 | +189% | 0 | 0 | — |
case-08 | pass→pass | 7,193 | 5,043 | -30% | 1 | 1 | 0% | 1,144 | 1,725 | +51% | 0 | 0 | — |
case-09 | pass→pass | 7,020 | 5,415 | -23% | 1 | 1 | 0% | 1,127 | 1,777 | +58% | 0 | 0 | — |
case-10 | fail→pass | 6,530 | 5,543 | -15% | 1 | 1 | 0% | 996 | 1,791 | +80% | 0 | 0 | — |
case-11 | pass→pass | 4,671 | 4,540 | -3% | 1 | 1 | 0% | 674 | 1,519 | +125% | 0 | 0 | — |
case-12 | pass→pass | 2,364 | 2,889 | +22% | 1 | 1 | 0% | 308 | 1,209 | +293% | 0 | 0 | — |
case-13 | fail→pass | 8,846 | 3,335 | -62% | 1 | 1 | 0% | 1,365 | 1,347 | -1% | 0 | 0 | — |
case-14 | fail→pass | 4,809 | 4,080 | -15% | 1 | 1 | 0% | 743 | 1,524 | +105% | 0 | 0 | — |
case-15 | pass→pass | 6,436 | 2,749 | -57% | 1 | 1 | 0% | 1,015 | 1,214 | +20% | 0 | 0 | — |
case-16 | pass→pass | 5,538 | 4,149 | -25% | 1 | 1 | 0% | 817 | 1,520 | +86% | 0 | 0 | — |
case-17 | pass→pass | 5,860 | 4,608 | -21% | 1 | 1 | 0% | 831 | 1,526 | +84% | 0 | 0 | — |
case-18 | pass→pass | 8,007 | 5,255 | -34% | 1 | 1 | 0% | 1,148 | 1,751 | +53% | 0 | 0 | — |
case-19 | pass→pass | 10,255 | 9,181 | -10% | 1 | 1 | 0% | 1,416 | 2,228 | +57% | 0 | 0 | — |
case-20 | pass→fail | 6,592 | 5,075 | -23% | 1 | 1 | 0% | 1,005 | 1,568 | +56% | 0 | 0 | — |
case-21 | pass→pass | 3,368 | 4,001 | +19% | 1 | 1 | 0% | 434 | 1,396 | +222% | 0 | 0 | — |
case-22 | pass→pass | 11,938 | 5,816 | -51% | 1 | 1 | 0% | 802 | 1,623 | +102% | 0 | 0 | — |
case-23 | pass→pass | 5,243 | 4,935 | -6% | 1 | 1 | 0% | 805 | 1,647 | +105% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +22 percentage points is the difference between those two pass rates over the 23 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.