Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when the user wants an inbox cleanup summary, morning email triage, scheduled inbox sweep, or a concise action-oriented Gmail digest delivered to chat.
.claude/skills/mkurman-inbox-triage/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -27% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 35% | 0% |
| case-16 | ✓→✗ | ▼ Worse | 2% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | -11% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | 1% | 0% |
Use this skill to turn a Gmail inbox scan into a short, actionable triage summary.
Gmail inbox triage is a two-step retrieval flow.
Core Gmail calls:
json{"plugin_name": "gmail", "endpoint_name": "list_inbox", "params": {"max_results": 10}}
Then fetch message details:
json{"plugin_name": "gmail", "endpoint_name": "get_message", "params": {"message_id": "ID_HERE"}}
Optional filtered search:
json{"plugin_name": "gmail", "endpoint_name": "search_messages", "params": {"query": "is:unread", "max_results": 10}}
Summarize messages into these buckets when possible:
If certainty is low, label the bucket as inferred.
Keep the result concise and operational.
Example outline:
textInbox triage Messages scanned: 10 Needs reply: 3 Needs action: 2 Can archive: 4 High priority: 1 Top actions: - Reply to Alice about project timeline - Review document request from Bob - Archive status update thread from ops
Where useful, recommend follow-up actions such as:
Use Gmail plugin actions only when the user explicitly asks to modify mailbox state.
For chat delivery, send the final concise summary through one of:
send_slack_messagesend_telegram_messagesend_discord_messageOnly deliver after the summary is readable and skim-friendly.
This skill is routine-friendly.
Typical trigger windows:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | pass→pass | 16,833 | 11,659 | -31% | 1 | 1 | 0% | 3,008 | 2,684 | -11% | 0 | 0 | — |
case-06 | pass→pass | 12,851 | 10,135 | -21% | 1 | 1 | 0% | 2,643 | 2,679 | +1% | 0 | 0 | — |
case-04 | pass→pass | 6,766 | 4,004 | -41% | 1 | 1 | 0% | 1,137 | 1,203 | +6% | 0 | 0 | — |
case-01 | fail→fail | 7,356 | 5,292 | -28% | 1 | 1 | 0% | 1,106 | 941 | -15% | 0 | 0 | — |
case-02 | fail→fail | 4,961 | 5,490 | +11% | 1 | 1 | 0% | 795 | 963 | +21% | 0 | 0 | — |
case-03 | fail→fail | 5,387 | 2,978 | -45% | 1 | 1 | 0% | 793 | 898 | +13% | 0 | 0 | — |
case-07 | fail→fail | 6,304 | 3,074 | -51% | 1 | 1 | 0% | 379 | 860 | +127% | 0 | 0 | — |
case-08 | fail→fail | 6,367 | 4,668 | -27% | 1 | 1 | 0% | 987 | 871 | -12% | 0 | 0 | — |
case-09 | fail→pass | 8,900 | 3,057 | -66% | 1 | 1 | 0% | 1,540 | 1,130 | -27% | 0 | 0 | — |
case-10 | fail→fail | 11,786 | 4,029 | -66% | 1 | 1 | 0% | 413 | 1,026 | +148% | 0 | 0 | — |
case-11 | fail→fail | 5,451 | 3,997 | -27% | 1 | 1 | 0% | 868 | 969 | +12% | 0 | 0 | — |
case-12 | fail→fail | 6,235 | 3,055 | -51% | 1 | 1 | 0% | 1,047 | 862 | -18% | 0 | 0 | — |
case-13 | pass→pass | 2,020 | 1,599 | -21% | 1 | 1 | 0% | 251 | 888 | +254% | 0 | 0 | — |
case-14 | fail→fail | 2,842 | 5,421 | +91% | 1 | 1 | 0% | 418 | 1,547 | +270% | 0 | 0 | — |
case-15 | fail→fail | 2,671 | 4,746 | +78% | 1 | 1 | 0% | 407 | 938 | +130% | 0 | 0 | — |
case-16 | pass→fail | 7,111 | 2,693 | -62% | 1 | 1 | 0% | 1,065 | 1,084 | +2% | 0 | 0 | — |
case-17 | fail→fail | 3,010 | 3,182 | +6% | 1 | 1 | 0% | 437 | 850 | +95% | 0 | 0 | — |
case-18 | fail→fail | 6,955 | 6,688 | -4% | 1 | 1 | 0% | 903 | 1,400 | +55% | 0 | 0 | — |
case-19 | fail→fail | 4,559 | 3,897 | -15% | 1 | 1 | 0% | 707 | 887 | +25% | 0 | 0 | — |
case-20 | fail→pass | 6,007 | 4,655 | -23% | 1 | 1 | 0% | 1,022 | 1,377 | +35% | 0 | 0 | — |
case-21 | fail→fail | 11,362 | 2,656 | -77% | 1 | 1 | 0% | 919 | 1,093 | +19% | 0 | 0 | — |
case-22 | fail→fail | 9,001 | 7,947 | -12% | 1 | 1 | 0% | 1,398 | 1,786 | +28% | 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, and 10 counted toward the lift figure. The other 12 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +5 percentage points is the difference between those two pass rates over the 10 comparable cases. 1 case got worse with the skill loaded, and it is 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.