Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Tracks per-agent token usage and flags waste in parallel dispatch. Use when evaluating parallel agent efficiency or after a multi-agent run.
.claude/skills/athola-agent-expenditure/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 60% | 0% |
Dispatching more agents does not always help. Coordination overhead grows with agent count:
| Agent Count | Expected Overhead | Guidance | |-------------|-------------------|----------| | 1-3 | Negligible | Dispatch freely | | 4-5 | 10-15% | Acceptable; plan first | | 6-8 | 20-30% | Monitor closely | | 9+ | 30%+ | Likely counterproductive |
Coordination overhead is measured as shared-file conflicts: concurrent Read/Write operations on the same file by different agents, as a percentage of total agent runtime.
After parallel agent runs, evaluate:
If 2+ questions answer no, reduce agent count in future dispatches of the same type.
See modules/waste-signals.md for the 5 waste signal categories and detection criteria.
.claude/rules/plan-before-large-dispatch.md for the 4+ agentplanning requirement
conserve:token-conservation for session-level token budgetingconjure:agent-teams for dispatch coordinationyes/no per agent (unique findings, proportional expenditure, no duplication, fewer agents sufficient)
modules/waste-signals.md checked againstthe completed run; any triggered signal named with the category
or "dispatch was efficient" with coordination overhead percentage
reduction is stated for future dispatches of the same type
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | fail→pass | 15,722 | 6,959 | -56% | 1 | 1 | 0% | 2,440 | 1,824 | -25% | 0 | 0 | — |
case-01 | fail→pass | 18,996 | 12,435 | -35% | 1 | 1 | 0% | 2,967 | 2,745 | -7% | 0 | 0 | — |
case-02 | fail→pass | 24,861 | 16,282 | -35% | 1 | 1 | 0% | 4,403 | 3,342 | -24% | 0 | 0 | — |
case-03 | fail→pass | 24,301 | 14,458 | -41% | 1 | 1 | 0% | 4,223 | 3,187 | -25% | 0 | 0 | — |
case-04 | fail→fail | 16,820 | 9,143 | -46% | 1 | 1 | 0% | 2,598 | 1,900 | -27% | 0 | 0 | — |
case-05 | fail→pass | 23,140 | 8,350 | -64% | 1 | 1 | 0% | 1,146 | 1,837 | +60% | 0 | 0 | — |
case-06 | fail→pass | 36,186 | 10,336 | -71% | 1 | 1 | 0% | 1,321 | 2,246 | +70% | 0 | 0 | — |
case-07 | fail→pass | 17,570 | 3,861 | -78% | 1 | 1 | 0% | 2,669 | 1,157 | -57% | 0 | 0 | — |
case-08 | fail→pass | 16,273 | 5,528 | -66% | 1 | 1 | 0% | 2,522 | 1,461 | -42% | 0 | 0 | — |
case-10 | pass→pass | 12,577 | 3,778 | -70% | 1 | 1 | 0% | 1,917 | 1,235 | -36% | 0 | 0 | — |
case-11 | fail→pass | 17,686 | 7,346 | -58% | 1 | 1 | 0% | 2,894 | 1,748 | -40% | 0 | 0 | — |
case-12 | fail→fail | 7,593 | 5,392 | -29% | 1 | 1 | 0% | 1,226 | 1,482 | +21% | 0 | 0 | — |
case-13 | pass→pass | 12,283 | 4,371 | -64% | 1 | 1 | 0% | 1,753 | 1,203 | -31% | 0 | 0 | — |
case-14 | fail→pass | 13,251 | 9,751 | -26% | 1 | 1 | 0% | 1,960 | 2,153 | +10% | 0 | 0 | — |
case-15 | pass→pass | 9,853 | 7,122 | -28% | 1 | 1 | 0% | 1,508 | 1,715 | +14% | 0 | 0 | — |
case-16 | pass→fail | 12,306 | 4,272 | -65% | 1 | 1 | 0% | 1,913 | 1,283 | -33% | 0 | 0 | — |
case-17 | pass→pass | 13,561 | 11,004 | -19% | 1 | 1 | 0% | 2,085 | 2,568 | +23% | 0 | 0 | — |
case-18 | pass→pass | 11,127 | 11,992 | +8% | 1 | 1 | 0% | 1,782 | 2,519 | +41% | 0 | 0 | — |
case-19 | pass→pass | 12,862 | 9,609 | -25% | 1 | 1 | 0% | 1,881 | 2,228 | +18% | 0 | 0 | — |
case-20 | pass→pass | 8,646 | 9,710 | +12% | 1 | 1 | 0% | 1,472 | 2,201 | +50% | 0 | 0 | — |
case-21 | fail→pass | 14,970 | 6,536 | -56% | 1 | 1 | 0% | 2,673 | 1,688 | -37% | 0 | 0 | — |
case-22 | pass→pass | 12,342 | 4,667 | -62% | 1 | 1 | 0% | 2,023 | 1,390 | -31% | 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 21 counted toward the lift figure. The other 1 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 +45 percentage points is the difference between those two pass rates over the 21 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.