Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Call a peer agent's A2A JSON-RPC endpoint. Supports tasks/send (send a task message) and tasks/get (query task status).
.claude/skills/ai-sns-a2a-json-rpc-call/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -52% | 0% |
This skill invokes A2A JSON-RPC methods on a remote agent endpoint.
Input parameters (via stdin JSON):
url (string, required): The peer's A2A JSON-RPC endpoint URL.method (string, required): "tasks/send" or "tasks/get".task_id (string, optional): Required for tasks/get. The task ID to query.message_text (string, optional): For tasks/send — a text message to include.message_data (object, optional): For tasks/send — a data payload to include.skill_id (string, optional): For tasks/send — the target skill ID on the peer agent.metadata (object, optional): Extra metadata to attach to the request.Output (stdout JSON):
ok (bool): Whether the call succeeded.result (object): The JSON-RPC result on success.error (string): Error message on failure.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | pass→pass | 16,803 | 14,493 | -14% | 1 | 1 | 0% | 3,309 | 3,169 | -4% | 0 | 0 | — |
case-04 | pass→fail | 12,627 | 8,107 | -36% | 1 | 1 | 0% | 2,640 | 1,677 | -36% | 0 | 0 | — |
case-06 | pass→pass | 10,860 | 10,070 | -7% | 1 | 1 | 0% | 2,174 | 2,260 | +4% | 0 | 0 | — |
case-07 | fail→pass | 7,409 | 3,324 | -55% | 1 | 1 | 0% | 1,538 | 905 | -41% | 0 | 0 | — |
case-01 | fail→fail | 6,863 | 6,424 | -6% | 1 | 1 | 0% | 1,391 | 540 | -61% | 0 | 0 | — |
case-02 | fail→fail | 13,767 | 13,608 | -1% | 1 | 1 | 0% | 863 | 474 | -45% | 0 | 0 | — |
case-03 | fail→fail | 5,899 | 5,199 | -12% | 1 | 1 | 0% | 1,239 | 554 | -55% | 0 | 0 | — |
case-08 | fail→pass | 7,332 | 4,157 | -43% | 1 | 1 | 0% | 1,443 | 1,022 | -29% | 0 | 0 | — |
case-09 | fail→pass | 7,620 | 3,111 | -59% | 1 | 1 | 0% | 1,481 | 724 | -51% | 0 | 0 | — |
case-10 | fail→pass | 8,456 | 3,984 | -53% | 1 | 1 | 0% | 1,735 | 942 | -46% | 0 | 0 | — |
case-11 | pass→pass | 5,058 | 4,107 | -19% | 1 | 1 | 0% | 1,009 | 1,054 | +4% | 0 | 0 | — |
case-12 | pass→pass | 4,116 | 3,281 | -20% | 1 | 1 | 0% | 750 | 570 | -24% | 0 | 0 | — |
case-13 | pass→pass | 2,808 | 6,909 | +146% | 1 | 1 | 0% | 519 | 911 | +76% | 0 | 0 | — |
case-14 | fail→pass | 7,293 | 2,271 | -69% | 1 | 1 | 0% | 1,420 | 684 | -52% | 0 | 0 | — |
case-15 | pass→pass | 5,842 | 3,147 | -46% | 1 | 1 | 0% | 1,150 | 752 | -35% | 0 | 0 | — |
case-16 | pass→pass | 31,838 | 2,774 | -91% | 1 | 1 | 0% | 6,125 | 694 | -89% | 0 | 0 | — |
case-17 | fail→pass | 5,865 | 2,279 | -61% | 1 | 1 | 0% | 1,100 | 596 | -46% | 0 | 0 | — |
case-18 | pass→pass | 5,178 | 2,738 | -47% | 1 | 1 | 0% | 964 | 781 | -19% | 0 | 0 | — |
case-19 | pass→pass | 7,992 | 1,503 | -81% | 1 | 1 | 0% | 1,669 | 484 | -71% | 0 | 0 | — |
case-20 | fail→pass | 9,260 | 4,181 | -55% | 1 | 1 | 0% | 1,785 | 733 | -59% | 0 | 0 | — |
case-21 | fail→pass | 6,098 | 3,291 | -46% | 1 | 1 | 0% | 1,132 | 819 | -28% | 0 | 0 | — |
case-22 | pass→pass | 8,943 | 2,566 | -71% | 1 | 1 | 0% | 1,663 | 687 | -59% | 0 | 0 | — |
case-23 | fail→pass | 5,000 | 2,954 | -41% | 1 | 1 | 0% | 1,008 | 785 | -22% | 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 +35 percentage points is the difference between those two pass rates over the 20 comparable cases. 2 cases got worse with the skill loaded, and they are 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.