Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Classify a batch of emails into action categories (reply now, reply later, archive, delete, unsubscribe) and surface unsubscribe candidates. Use after a busy week, running inbox zero, or triaging email overload.
.claude/skills/borghei-email-triage/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 44% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 388% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 113% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 79% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 5% | 0% |
Classify a batch of email subjects + senders into action buckets and surface inbox-zero candidates.
email, inbox, inbox zero, triage, unsubscribe, mailing list, mailbox, gmail, outlook, productivity
Before triaging, confirm these inputs. If any is unknown or vague, ASK — do not assume:
Stop rule: ask only the 2-3 that most change the output. If the user says "just draft it," proceed and list your assumptions at the top of the artifact.
subject,sender,snippet,received_atpython scripts/email_classifier.py inbox.csvassets/gmail_filter_template.md) so future similar emails route automaticallyTime Estimate: 30-45 minutes for a busy week.
Time Estimate: 15 minutes per pass.
references/inbox_zero_method.mdTime Estimate: 1-2 hours one-time; then 20 min/week to maintain.
Classifies email rows into action buckets using rule-based pattern matching on sender domain, subject line, and snippet.
bashpython scripts/email_classifier.py inbox.csv python scripts/email_classifier.py inbox.csv --json
Action buckets:
references/inbox_zero_method.md — Method, daily routine, common pitfallsassets/gmail_filter_template.md — Common Gmail filter recipes for the action buckets above| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 6,090 | 3,789 | -38% | 1 | 1 | 0% | 960 | 1,387 | +44% | 0 | 0 | — |
case-02 | fail→pass | 2,344 | 5,661 | +142% | 1 | 1 | 0% | 335 | 1,635 | +388% | 0 | 0 | — |
case-03 | fail→pass | 5,735 | 6,651 | +16% | 1 | 1 | 0% | 828 | 1,766 | +113% | 0 | 0 | — |
case-04 | fail→pass | 9,834 | 12,933 | +32% | 1 | 1 | 0% | 1,555 | 2,778 | +79% | 0 | 0 | — |
case-05 | fail→pass | 6,563 | 1,692 | -74% | 1 | 1 | 0% | 1,049 | 1,102 | +5% | 0 | 0 | — |
case-06 | fail→pass | 7,902 | 5,385 | -32% | 1 | 1 | 0% | 1,111 | 1,647 | +48% | 0 | 0 | — |
case-07 | pass→pass | 13,075 | 10,658 | -18% | 1 | 1 | 0% | 2,032 | 2,486 | +22% | 0 | 0 | — |
case-20 | fail→pass | 10,340 | 10,345 | +0% | 1 | 1 | 0% | 1,481 | 2,347 | +58% | 0 | 0 | — |
case-08 | pass→pass | 10,523 | 6,860 | -35% | 1 | 1 | 0% | 1,488 | 1,729 | +16% | 0 | 0 | — |
case-09 | pass→pass | 9,312 | 6,058 | -35% | 1 | 1 | 0% | 1,440 | 1,522 | +6% | 0 | 0 | — |
case-10 | fail→pass | 10,520 | 3,665 | -65% | 1 | 1 | 0% | 1,835 | 1,313 | -28% | 0 | 0 | — |
case-11 | pass→pass | 12,128 | 1,859 | -85% | 1 | 1 | 0% | 1,926 | 1,098 | -43% | 0 | 0 | — |
case-12 | pass→pass | 11,443 | 11,024 | -4% | 1 | 1 | 0% | 1,751 | 2,400 | +37% | 0 | 0 | — |
case-13 | fail→fail | 11,086 | 5,769 | -48% | 1 | 1 | 0% | 1,767 | 1,581 | -11% | 0 | 0 | — |
case-14 | pass→pass | 16,440 | 11,560 | -30% | 1 | 1 | 0% | 1,946 | 2,616 | +34% | 0 | 0 | — |
case-15 | fail→pass | 12,363 | 4,539 | -63% | 1 | 1 | 0% | 1,816 | 1,519 | -16% | 0 | 0 | — |
case-16 | pass→pass | 10,560 | 4,028 | -62% | 1 | 1 | 0% | 1,528 | 1,436 | -6% | 0 | 0 | — |
case-17 | pass→pass | 7,003 | 7,010 | +0% | 1 | 1 | 0% | 1,176 | 1,907 | +62% | 0 | 0 | — |
case-18 | fail→pass | 10,328 | 6,843 | -34% | 1 | 1 | 0% | 1,482 | 1,820 | +23% | 0 | 0 | — |
case-19 | pass→pass | 11,747 | 8,659 | -26% | 1 | 1 | 0% | 1,658 | 2,081 | +26% | 0 | 0 | — |
case-21 | pass→pass | 13,858 | 12,178 | -12% | 1 | 1 | 0% | 2,484 | 3,033 | +22% | 0 | 0 | — |
case-22 | pass→pass | 15,201 | 12,972 | -15% | 1 | 1 | 0% | 2,790 | 3,137 | +12% | 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. 22 cases were attempted. The headline lift of +45 percentage points is the difference between those two pass rates over the 22 comparable cases.
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.