RelayPlane

/** * relay_skills_list Tool * * List available pre-built workflow skills. * Skills are reusable patterns for common tasks. */ import { z } from 'zod'; export const relaySkillsListSchema = z.object({ category: z .enum(['extraction', 'content', 'integration', 'all']) .optional() .describe('Optional: filter by category'), }); export type RelaySkillsListInput = z.infer<typeof relaySkillsListSchema>; export interface SkillInfo { name: string; category: string; description: string; models: string[]; contextReduction: string; usage: string; estimatedCost: string; } // Pre-built skills registry const SKILLS: SkillInfo[] = [ { name: 'invoice-processor', category: 'extraction', description: 'Extract, validate, and summarize invoice data from images or PDFs.', models: ['openai:gpt-4o', 'anthropic:claude-3-5-sonnet-20241022', 'openai:gpt-4o-mini'], contextReduction: '97% (from ~15k to ~500 tokens)', usage: `await relay.workflow("invoice-processor") .step("extract").with("openai:gpt-4o").prompt("Extract invoice fields...") .step("validate").with("anthropic:claude-3-5-sonnet-20241022").depends("extract").prompt("Verify totals...") .step("summarize").with("openai:gpt-4o-mini").depends("validate").prompt("Create summary...") .run({ fileUrl: "..." })`, estimatedCost: '$0.02-0.05 per invoice', }, { name: 'content-pipeline', category: 'content', description: 'Research, draft, and refine content with multiple AI passes.', models: ['openai:gpt-4o', 'anthropic:claude-3-5-sonnet-20241022'], contextReduction: '85% (from ~20k to ~3k tokens)', usage: `await relay.workflow("content-pipeline") .step("research").with("openai:gpt-4o").prompt("Research topic: {{input.topic}}...") .step("draft").with("anthropic:claude-3-5-sonnet-20241022").depends("research").prompt("Write draft...") .step("refine").with("openai:gpt-4o").depends("draft").prompt("Polish and finalize...") .run({ topic: "..." })`, estimatedCost: '$0.05-0.15 per piece', }, { name: 'lead-enrichment', category: 'integration', description: 'Enrich lead data with company info, social profiles, and qualification scores.', models: ['openai:gpt-4o-mini', 'anthropic:claude-3-5-haiku-20241022'], contextReduction: '90% (from ~10k to ~1k tokens)', usage: `await relay.workflow("lead-enrichment") .step("lookup").mcp("clearbit:lookup").params({ email: "{{input.email}}" }) .step("analyze").with("anthropic:claude-3-5-haiku-20241022").depends("lookup").prompt("Score lead...") .step("format").with("openai:gpt-4o-mini").depends("analyze").prompt("Format for CRM...") .run({ email: "..." })`, estimatedCost: '$0.01-0.03 per lead', }, { name: 'document-qa', category: 'extraction', description: 'Answer questions about documents using RAG pattern with context preservation.', models: ['openai:gpt-4o', 'anthropic:claude-3-5-sonnet-20241022'], contextReduction: '75% (from ~50k to ~12k tokens)', usage: `await relay.workflow("document-qa") .step("chunk").with("openai:gpt-4o-mini").prompt("Split document into chunks...") .step("embed").mcp("embeddings:create").params({ texts: "{{steps.chunk.output}}" }) .step("answer").with("openai:gpt-4o").depends("embed").prompt("Answer: {{input.question}}...") .run({ document: "...", question: "..." })`, estimatedCost: '$0.02-0.10 per query', }, { name: 'code-review', category: 'content', description: 'Multi-stage code review: security, performance, and style analysis.', models: ['anthropic:claude-3-5-sonnet-20241022', 'openai:gpt-4o'], contextReduction: '80% (from ~25k to ~5k tokens)', usage: `await relay.workflow("code-review") .step("security").with("anthropic:claude-3-5-sonnet-20241022").prompt("Analyze security...") .step("performance").with("openai:gpt-4o").prompt("Analyze performance...") .step("style").with("openai:gpt-4o-mini").prompt("Check code style...") .step("summary").with("anthropic:claude-3-5-sonnet-20241022").depends("security", "performance", "style").prompt("Summarize findings...") .run({ code: "..." })`, estimatedCost: '$0.05-0.20 per review', }, { name: 'email-classifier', category: 'extraction', description: 'Classify emails by intent, urgency, and route to appropriate handlers.', models: ['openai:gpt-4o-mini', 'anthropic:claude-3-5-haiku-20241022'], contextReduction: '95% (from ~8k to ~400 tokens)', usage: `await relay.workflow("email-classifier") .step("classify").with("openai:gpt-4o-mini").prompt("Classify email intent and urgency...") .step("route").with("anthropic:claude-3-5-haiku-20241022").depends("classify").prompt("Determine routing...") .run({ email: "..." })`, estimatedCost: '$0.001-0.005 per email', }, ]; export interface RelaySkillsListResponse { skills: SkillInfo[]; } export async function relaySkillsList( input: RelaySkillsListInput ): Promise<RelaySkillsListResponse> { let skills = [...SKILLS]; // Filter by category if specified if (input.category && input.category !== 'all') { skills = skills.filter(s => s.category === input.category); } // Sort by category, then by name skills.sort((a, b) => { if (a.category !== b.category) { return a.category.localeCompare(b.category); } return a.name.localeCompare(b.name); }); return { skills }; } export const relaySkillsListDefinition = { name: 'relay_skills_list', description: 'List available pre-built workflow skills. Skills are reusable patterns for common tasks (invoice processing, content pipelines, etc.). Returns skill names, descriptions, context reduction metrics, and usage examples.', inputSchema: { type: 'object' as const, properties: { category: { type: 'string', description: 'Optional: filter by category', enum: ['extraction', 'content', 'integration', 'all'], }, }, }, };