Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Claw background job management using x agent run and x agent job. Use when the agent needs to offload complex or long-running tasks that would block the chat.
.claude/skills/x-cmd-claw-background-jobs/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -36% | 0% |
Use x agent run for tasks estimated >2 minutes or that would block the chat:
x agent run --job-id "<id>" --max-iterations <n> "<task>": Create and start an async job. The AI auto-generates a PLAN.md and iterates until done or max iterations reached.x agent job status --job-id "<id>" --llms: Check progress, iteration count, and whether the job is active/completed (YAML output for parsing).x agent job ls --active --llms: List all active jobs.x agent job stop --job-id "<id>": Stop the background process.Use <im>-<chatid>-<brief> for traceability, e.g., weixin-user123-loganalysis.
x agent run executes in a fresh environment with NO chat memory. The <task> must be self-contained: clear goal, required tools, expected output.x claw agentrequest to report results.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 28,045 | 22,639 | -19% | 1 | 1 | 0% | 1,828 | 1,087 | -41% | 0 | 0 | — |
case-02 | fail→fail | 5,471 | 7,688 | +41% | 1 | 1 | 0% | 754 | 802 | +6% | 0 | 0 | — |
case-03 | fail→pass | 11,079 | 11,009 | -1% | 1 | 1 | 0% | 797 | 837 | +5% | 0 | 0 | — |
case-04 | pass→pass | 4,534 | 3,668 | -19% | 1 | 1 | 0% | 433 | 696 | +61% | 0 | 0 | — |
case-05 | pass→pass | 15,825 | 7,461 | -53% | 1 | 1 | 0% | 939 | 1,216 | +29% | 0 | 0 | — |
case-06 | pass→fail | 3,504 | 7,727 | +121% | 1 | 1 | 0% | 438 | 1,272 | +190% | 0 | 0 | — |
case-07 | fail→pass | 9,950 | 5,273 | -47% | 1 | 1 | 0% | 1,346 | 921 | -32% | 0 | 0 | — |
case-08 | fail→pass | 9,419 | 5,757 | -39% | 1 | 1 | 0% | 1,250 | 1,028 | -18% | 0 | 0 | — |
case-09 | fail→pass | 5,131 | 2,437 | -53% | 1 | 1 | 0% | 723 | 461 | -36% | 0 | 0 | — |
case-10 | fail→pass | 8,624 | 3,384 | -61% | 1 | 1 | 0% | 1,033 | 561 | -46% | 0 | 0 | — |
case-11 | fail→pass | 4,359 | 2,194 | -50% | 1 | 1 | 0% | 701 | 515 | -27% | 0 | 0 | — |
case-12 | fail→fail | 6,015 | 6,348 | +6% | 1 | 1 | 0% | 623 | 1,341 | +115% | 0 | 0 | — |
case-13 | fail→pass | 9,681 | 3,468 | -64% | 1 | 1 | 0% | 1,143 | 779 | -32% | 0 | 0 | — |
case-14 | fail→pass | 12,574 | 4,856 | -61% | 1 | 1 | 0% | 1,871 | 827 | -56% | 0 | 0 | — |
case-15 | fail→pass | 13,034 | 3,826 | -71% | 1 | 1 | 0% | 1,909 | 825 | -57% | 0 | 0 | — |
case-16 | fail→pass | 12,052 | 4,959 | -59% | 1 | 1 | 0% | 1,605 | 1,042 | -35% | 0 | 0 | — |
case-17 | fail→pass | 5,204 | 2,524 | -51% | 1 | 1 | 0% | 591 | 591 | 0% | 0 | 0 | — |
case-18 | fail→pass | 9,586 | 2,521 | -74% | 1 | 1 | 0% | 1,487 | 540 | -64% | 0 | 0 | — |
case-19 | fail→pass | 8,628 | 5,802 | -33% | 1 | 1 | 0% | 1,164 | 735 | -37% | 0 | 0 | — |
case-20 | fail→pass | 53,128 | 14,574 | -73% | 1 | 1 | 0% | 8,219 | 2,774 | -66% | 0 | 0 | — |
case-21 | fail→fail | 9,395 | 7,832 | -17% | 1 | 1 | 0% | 258 | 711 | +176% | 0 | 0 | — |
case-22 | pass→pass | 8,799 | 6,402 | -27% | 1 | 1 | 0% | 1,297 | 1,060 | -18% | 0 | 0 | — |
case-23 | pass→pass | 7,495 | 6,547 | -13% | 1 | 1 | 0% | 1,159 | 1,224 | +6% | 0 | 0 | — |
case-24 | fail→pass | 10,770 | 3,067 | -72% | 1 | 1 | 0% | 1,618 | 706 | -56% | 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. 24 cases were attempted, and 22 counted toward the lift figure. The other 2 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 +63 percentage points is the difference between those two pass rates over the 22 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.