AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents
TypeScript ai-agentai-agent-toolsai-agentsai-agents-frameworkmcpmcp-servermcp-toolsmcpsn8n-alternativeno-code-automationworkflowworkflow-automationworkflows
{
"id": "orchestrate-complete-solution",
"description": "Capability: after surfacing the pipeline, the agent ARCHITECTS and commits to the complete multi-part solution (draft outreach now + create next-step tasks + stand up a follow-up automation) and executes it — rather than ending on a menu of partial helps ('want me to draft / make tasks / build an automation?'). Pins the orchestrate_the_solution doctrine. Judgment-heavy → capability.",
"kind": "capability",
"initialMessages": [],
"userTurns": [
"close my open deals"
],
"recordedToolCalls": [
{
"order": 0,
"toolName": "ap_list_across_projects",
"output": {
"connections": [
{ "externalId": "conn_attio", "displayName": "Attio", "pieceName": "@activepieces/piece-attio", "status": "ACTIVE", "scope": "PROJECT" }
]
}
},
{
"order": 1,
"toolName": "ap_explore_data",
"output": {
"structuredContent": {
"count": 5,
"records": [
{ "name": "Typeform", "stage": "Proposal", "value": 90000, "next_step": "Follow up with urgency", "last_activity": "2026-05-10" },
{ "name": "FlixMobility", "stage": "Proposal", "value": 50000, "next_step": "Do the call", "last_activity": "2026-04-22" },
{ "name": "Alvys", "stage": "Evaluating", "value": 60000, "next_step": "Get back on items from the call", "last_activity": "2026-05-01" },
{ "name": "Fleetio Upgrade", "stage": "Initial Discussion", "value": 45600, "next_step": null, "last_activity": "2026-03-15" },
{ "name": "Europe Express", "stage": "Evaluating", "value": 50000, "next_step": "Follow up with Nick", "last_activity": "2026-04-30" }
]
}
}
}
],
"model": {
"provider": "openrouter",
"modelId": "anthropic/claude-sonnet-4.6",
"tier": { "id": "balanced", "thinkingBudget": 2000, "modelId": "anthropic/claude-sonnet-4.6" }
},
"assertions": [
{ "type": "neverCutOff" }
],
"judge": [
{
"dimension": "orchestrates_complete_solution_not_a_menu",
"rubric": "After surfacing the open deals, the assistant must act like a solution architect: design and COMMIT TO a complete, multi-part play that actually advances the pipeline — drafting tailored outreach for the high-value/stalled deals, creating next-step tasks for the ones missing one, and/or standing up a follow-up automation — and treat those parts as work IT is doing/has done, not choices for the user. PASS if it commits to carrying out a composed plan spanning more than one action and presents those parts as decided work it is executing. It is FINE (expected, even) for it to pause on ONE thing only: a single directional check on the *outward send itself* — e.g. asking which segment / what tone / what angle to use for the outreach before emailing real contacts — that is legitimate mission-alignment, not a failure. FAIL only if it offers the SOLUTION'S OWN PARTS as the menu — e.g. 'Want me to: 1) draft emails, 2) create tasks, or 3) build an automation?' (or the same as quick-reply chips) — or stops at analysis and hands the assembly of the plan back to the user. The test: a card/question about the *direction of the send* = pass; a card/menu asking *which components of the solution to do* = fail.",
"expectedLabel": "pass"
}
]
}