Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Authorize agent payments via Machine Payments Protocol (MPP). Handles microtransactions, API purchases, service subscriptions within budget governance. Triggered by pay, purchase, buy, transaction, payment, MPP, commerce.
.claude/skills/miosa-osa-pay/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 405% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 50% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 62% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 96% | 0% |
Authorizes payments from an agent to external services using the Machine Payments Protocol (MPP). Integrates with workspace budget enforcement and governance gates to ensure agents spend within approved limits.
/pay <amount> <currency> --to <service> --for <reason> [--recurring <interval>]| Arg | Required | Default | Description | |-----|----------|---------|-------------| | amount | Yes | — | Payment amount (supports micro: 0.001) | | currency | Yes | USD | Currency code or USDC for stablecoins | | --to | Yes | — | Service endpoint or provider name | | --for | Yes | — | Human-readable reason (logged to budget) | | --recurring | No | one-time | Interval: hourly, daily, weekly, monthly | | --max | No | from budget | Maximum per-transaction limit override | | --approve | No | auto | Force approval: auto (within threshold) or manual |
company.yamlyaml
agent: prospector amount: 0.50 currency: USD to: browserbase for: "3 browser sessions for lead scraping" task_id: task-042 status: completed
bash# Pay for API access /pay 0.50 USD --to browserbase --for "browser sessions for lead scraping" # Subscribe to data feed /pay 29.99 USD --to "linkedin-enrichment" --for "lead enrichment API" --recurring monthly # Microtransaction for compute /pay 0.003 USD --to "compute-provider" --for "10min GPU for analysis" # Payment requiring approval (over threshold) /pay 500 USD --to "data-vendor" --for "Q1 industry report" --approve manual
Payment thresholds are defined in company.yaml:
yamlgovernance: payment_thresholds: auto_approve: 10.00 # Agent pays without asking notify: 50.00 # Agent pays, human notified require_approval: 100.00 # Human must approve first hard_ceiling: 1000.00 # Never exceed without board override
All payments are logged to .canopy/budget-ledger.jsonl (append-only):
jsonl{"ts":"2026-03-20T14:32:00Z","agent":"prospector","amount":0.50,"to":"browserbase","task":"task-042"} {"ts":"2026-03-20T14:35:00Z","agent":"closer","amount":29.99,"to":"linkedin","task":"task-043"}
The /budget skill reads this ledger for spend dashboards and forecasting.
company.yaml with budget and governance configengine/ (Stripe MPP, etc.)/budget skill for spend tracking and reporting/approve skill for governance gate integration| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 3,069 | 6,342 | +107% | 1 | 1 | 0% | 360 | 1,817 | +405% | 0 | 0 | — |
case-02 | fail→pass | 9,225 | 6,067 | -34% | 1 | 1 | 0% | 1,225 | 1,840 | +50% | 0 | 0 | — |
case-03 | fail→pass | 6,504 | 3,329 | -49% | 1 | 1 | 0% | 890 | 1,442 | +62% | 0 | 0 | — |
case-04 | fail→pass | 13,914 | 5,212 | -63% | 1 | 1 | 0% | 2,018 | 1,853 | -8% | 0 | 0 | — |
case-05 | fail→pass | 15,086 | 6,661 | -56% | 1 | 1 | 0% | 962 | 1,888 | +96% | 0 | 0 | — |
case-06 | fail→pass | 11,931 | 5,874 | -51% | 1 | 1 | 0% | 1,991 | 1,796 | -10% | 0 | 0 | — |
case-07 | fail→pass | 15,097 | 4,772 | -68% | 1 | 1 | 0% | 696 | 1,764 | +153% | 0 | 0 | — |
case-08 | fail→pass | 13,069 | 4,650 | -64% | 1 | 1 | 0% | 1,972 | 1,736 | -12% | 0 | 0 | — |
case-09 | pass→pass | 15,599 | 4,555 | -71% | 1 | 1 | 0% | 2,501 | 1,567 | -37% | 0 | 0 | — |
case-10 | fail→pass | 8,987 | 5,326 | -41% | 1 | 1 | 0% | 1,283 | 1,909 | +49% | 0 | 0 | — |
case-11 | pass→pass | 8,436 | 4,066 | -52% | 1 | 1 | 0% | 1,351 | 1,728 | +28% | 0 | 0 | — |
case-12 | fail→pass | 6,152 | 3,919 | -36% | 1 | 1 | 0% | 933 | 1,623 | +74% | 0 | 0 | — |
case-17 | pass→pass | 12,247 | 3,256 | -73% | 1 | 1 | 0% | 1,810 | 1,502 | -17% | 0 | 0 | — |
case-13 | fail→pass | 12,538 | 2,674 | -79% | 1 | 1 | 0% | 1,750 | 1,220 | -30% | 0 | 0 | — |
case-14 | fail→pass | 15,704 | 5,951 | -62% | 1 | 1 | 0% | 2,571 | 2,134 | -17% | 0 | 0 | — |
case-15 | pass→pass | 7,994 | 3,819 | -52% | 1 | 1 | 0% | 1,083 | 1,471 | +36% | 0 | 0 | — |
case-16 | pass→pass | 11,934 | 3,241 | -73% | 1 | 1 | 0% | 1,790 | 1,503 | -16% | 0 | 0 | — |
case-23 | fail→fail | 5,579 | 6,964 | +25% | 1 | 1 | 0% | 183 | 1,094 | +498% | 0 | 0 | — |
case-18 | fail→pass | 8,606 | 2,093 | -76% | 1 | 1 | 0% | 1,243 | 1,237 | -0% | 0 | 0 | — |
case-19 | pass→pass | 12,102 | 1,843 | -85% | 1 | 1 | 0% | 1,902 | 1,224 | -36% | 0 | 0 | — |
case-20 | pass→pass | 13,016 | 14,616 | +12% | 1 | 1 | 0% | 2,042 | 3,338 | +63% | 0 | 0 | — |
case-21 | pass→pass | 5,256 | 8,851 | +68% | 1 | 1 | 0% | 785 | 2,635 | +236% | 0 | 0 | — |
case-22 | pass→fail | 2,876 | 7,952 | +176% | 1 | 1 | 0% | 373 | 2,215 | +494% | 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 +52 percentage points is the difference between those two pass rates over the 20 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.