Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Turn a recurring chore into a Superset automation — drafts the agent prompt, confirms schedule and target, creates it with the CLI, and reviews the first run together. Use when the user wants a scheduled or recurring agent, a daily/weekly job, or to automate a repeating task with Superset.
.claude/skills/superset-sh-automate/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 107% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -42% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -54% | 0% |
Turn "I keep doing X every morning" into an automation that does X on a schedule.
Pin down: the outcome, the cadence, the inputs it reads, and what "done" looks like. Then draft the automation prompt — write it as instructions for an agent with zero context, and if the task has rules that will evolve (triage criteria, formats), put them in a document the automation reads at runtime so they can be edited without touching the prompt.
superset projects list — a project target creates a fresh workspace per run (most tasks)superset workspaces list — a workspace target reuses the same workspace every run (stateful tasks)Show the user (use the ask_user tool if available): the name, the schedule as an RRULE, the agent, and the target — plus the exact command you will run. Never create without explicit confirmation.
bashsuperset automations create \ --name "Daily issue triage" \ --rrule "FREQ=DAILY;BYHOUR=9;BYMINUTE=0" \ --timezone America/Los_Angeles \ --project <id> \ --agent claude \ --prompt-file /tmp/automation-prompt.md
(--workspace <id> instead of --project for reuse mode; --host <id> if it should run on another machine; prefer --prompt-file for multiline prompts.)
Then trigger a first run now with superset automations run <id>, review superset automations logs <id> with the user, and refine the prompt via superset automations prompt set <id> until the run output is right. An automation isn't done until one real run looked good.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 10,887 | 6,446 | -41% | 1 | 1 | 0% | 1,807 | 833 | -54% | 0 | 0 | — |
case-02 | fail→fail | 15,543 | 17,100 | +10% | 1 | 1 | 0% | 2,584 | 997 | -61% | 0 | 0 | — |
case-03 | fail→fail | 12,440 | 5,222 | -58% | 1 | 1 | 0% | 2,068 | 1,244 | -40% | 0 | 0 | — |
case-04 | pass→pass | 12,814 | 9,596 | -25% | 1 | 1 | 0% | 1,956 | 2,049 | +5% | 0 | 0 | — |
case-05 | fail→pass | 7,900 | 3,147 | -60% | 1 | 1 | 0% | 1,333 | 983 | -26% | 0 | 0 | — |
case-10 | fail→pass | 8,767 | 2,966 | -66% | 1 | 1 | 0% | 1,438 | 935 | -35% | 0 | 0 | — |
case-06 | pass→pass | 18,738 | 3,220 | -83% | 1 | 1 | 0% | 2,589 | 882 | -66% | 0 | 0 | — |
case-07 | fail→pass | 8,437 | 9,623 | +14% | 1 | 1 | 0% | 913 | 1,893 | +107% | 0 | 0 | — |
case-08 | fail→pass | 11,382 | 3,646 | -68% | 1 | 1 | 0% | 1,864 | 1,072 | -42% | 0 | 0 | — |
case-09 | pass→pass | 9,697 | 3,724 | -62% | 1 | 1 | 0% | 1,509 | 1,079 | -28% | 0 | 0 | — |
case-11 | pass→pass | 12,710 | 2,448 | -81% | 1 | 1 | 0% | 2,133 | 827 | -61% | 0 | 0 | — |
case-12 | fail→fail | 8,469 | 3,287 | -61% | 1 | 1 | 0% | 1,506 | 1,091 | -28% | 0 | 0 | — |
case-13 | fail→fail | 17,452 | 5,176 | -70% | 1 | 1 | 0% | 906 | 1,425 | +57% | 0 | 0 | — |
case-14 | pass→pass | 12,299 | 7,557 | -39% | 1 | 1 | 0% | 1,883 | 1,549 | -18% | 0 | 0 | — |
case-15 | pass→pass | 9,689 | 1,904 | -80% | 1 | 1 | 0% | 1,608 | 738 | -54% | 0 | 0 | — |
case-16 | fail→pass | 10,489 | 2,449 | -77% | 1 | 1 | 0% | 1,793 | 829 | -54% | 0 | 0 | — |
case-17 | pass→pass | 6,518 | 3,500 | -46% | 1 | 1 | 0% | 1,006 | 1,004 | -0% | 0 | 0 | — |
case-18 | pass→pass | 8,223 | 2,758 | -66% | 1 | 1 | 0% | 1,242 | 893 | -28% | 0 | 0 | — |
case-19 | fail→pass | 9,531 | 2,594 | -73% | 1 | 1 | 0% | 1,598 | 848 | -47% | 0 | 0 | — |
case-20 | pass→pass | 5,256 | 3,026 | -42% | 1 | 1 | 0% | 827 | 932 | +13% | 0 | 0 | — |
case-21 | fail→fail | 12,883 | 9,850 | -24% | 1 | 1 | 0% | 2,208 | 2,187 | -1% | 0 | 0 | — |
case-22 | pass→pass | 12,957 | 7,569 | -42% | 1 | 1 | 0% | 2,083 | 1,798 | -14% | 0 | 0 | — |
case-23 | pass→pass | 8,226 | 3,142 | -62% | 1 | 1 | 0% | 1,047 | 896 | -14% | 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, and 20 counted toward the lift figure. The other 3 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 +26 percentage points is the difference between those two pass rates over the 20 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.