Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →An advanced skill for L3 autonomous loops. When the token budget nears exhaustion, the agent analyzes its ROI and autonomously drafts a negotiation request for a budget increase rather than silently failing.
.claude/skills/cobusgreyling-budget-negotiator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-20 | ✗→✓ | ▲ Improved | 84% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 40% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 5% | 0% |
<!-- NOTE: This template is a mirror of skills/budget-negotiator/SKILL.md. Any changes made here must be kept byte-identical to the source skill. -->
When you detect that you are nearing your daily token cap (e.g., ≥90% spend) via the loop-budget skill, DO NOT just exit silently. You must negotiate for an extension based on your Return on Investment (ROI).
Ensure that critical work is not arbitrarily blocked by token budgets if the loop has demonstrated a high success rate and the remaining work is critical.
budget-negotiator ONLY activates when loop-budget would otherwise hard-exit (i.e. ≥90% spend), and ONLY when there are High Priority items remaining. If work is not High Priority, yield to the standard loop-budget exit sequence.loop-budget.md. A human MUST explicitly edit loop-budget.md to increase the cap before you can resume.[BUDGET NEGOTIATION] tag without increasing the budget, you must revert to report-only mode and NOT ask again today.loop-run-log.md and sum up the actions_taken, outcome (specifically successes), and tokens_estimate fields for today.STATE.md to identify the severity of the remaining actionable items in the Watch List or High Priority sections.actions_taken count and successful outcomes you've completed today based on the JSON log.loop-budget.md field (e.g., "Requesting +50k tokens for the 'Validate/Audit (CI)' max tokens/day").STATE.md with the tag [BUDGET NEGOTIATION].[BUDGET NEGOTIATION] I have burned 95k/100k tokens for the 'Daily Triage' loop today. However, I have successfully executed 4 actions with a 'fix-proposed' outcome. There is 1 critical CI failure remaining on the main branch. I request a temporary bump of 'Daily Triage' max tokens/day by +50k to resolve it.
WAITING_FOR_BUDGET and safely suspend execution.loop-budget.md on a subsequent run and verify that the human has increased your token cap.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-20 | fail→pass | 18,574 | 5,163 | -72% | 1 | 1 | 0% | 953 | 1,754 | +84% | 0 | 0 | — |
case-01 | fail→fail | 16,217 | 7,473 | -54% | 1 | 1 | 0% | 2,335 | 1,349 | -42% | 0 | 0 | — |
case-02 | fail→fail | 12,520 | 6,511 | -48% | 1 | 1 | 0% | 1,699 | 1,351 | -20% | 0 | 0 | — |
case-03 | fail→fail | 4,582 | 4,321 | -6% | 1 | 1 | 0% | 221 | 1,210 | +448% | 0 | 0 | — |
case-04 | pass→pass | 3,670 | 2,428 | -34% | 1 | 1 | 0% | 625 | 1,323 | +112% | 0 | 0 | — |
case-05 | fail→pass | 7,496 | 5,551 | -26% | 1 | 1 | 0% | 1,325 | 1,851 | +40% | 0 | 0 | — |
case-06 | fail→pass | 7,742 | 3,572 | -54% | 1 | 1 | 0% | 1,332 | 1,358 | +2% | 0 | 0 | — |
case-07 | fail→pass | 10,770 | 2,964 | -72% | 1 | 1 | 0% | 1,877 | 1,422 | -24% | 0 | 0 | — |
case-08 | pass→pass | 8,145 | 4,172 | -49% | 1 | 1 | 0% | 1,400 | 1,640 | +17% | 0 | 0 | — |
case-09 | fail→pass | 8,474 | 3,478 | -59% | 1 | 1 | 0% | 1,347 | 1,416 | +5% | 0 | 0 | — |
case-10 | fail→pass | 9,043 | 3,368 | -63% | 1 | 1 | 0% | 1,422 | 1,327 | -7% | 0 | 0 | — |
case-11 | fail→pass | 12,336 | 2,334 | -81% | 1 | 1 | 0% | 2,373 | 1,253 | -47% | 0 | 0 | — |
case-12 | fail→pass | 9,402 | 2,039 | -78% | 1 | 1 | 0% | 1,783 | 1,126 | -37% | 0 | 0 | — |
case-13 | pass→pass | 8,744 | 4,265 | -51% | 1 | 1 | 0% | 1,469 | 1,669 | +14% | 0 | 0 | — |
case-14 | fail→pass | 8,686 | 5,468 | -37% | 1 | 1 | 0% | 1,432 | 1,868 | +30% | 0 | 0 | — |
case-15 | fail→pass | 11,624 | 3,834 | -67% | 1 | 1 | 0% | 1,814 | 1,539 | -15% | 0 | 0 | — |
case-16 | fail→fail | 10,107 | 2,221 | -78% | 1 | 1 | 0% | 1,665 | 1,239 | -26% | 0 | 0 | — |
case-17 | fail→pass | 5,066 | 3,258 | -36% | 1 | 1 | 0% | 834 | 1,414 | +70% | 0 | 0 | — |
case-18 | fail→pass | 4,673 | 3,641 | -22% | 1 | 1 | 0% | 778 | 1,512 | +94% | 0 | 0 | — |
case-19 | pass→pass | 6,343 | 3,944 | -38% | 1 | 1 | 0% | 1,054 | 1,494 | +42% | 0 | 0 | — |
case-21 | pass→pass | 14,496 | 7,698 | -47% | 1 | 1 | 0% | 2,852 | 2,458 | -14% | 0 | 0 | — |
case-22 | fail→pass | 8,067 | 1,766 | -78% | 1 | 1 | 0% | 564 | 1,121 | +99% | 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 17 counted toward the lift figure. The other 5 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 +59 percentage points is the difference between those two pass rates over the 17 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.