RelayPlane Logo

Skills

Pre-built workflow patterns with documented context reduction metrics.

Skills are reusable workflow patterns that solve common tasks. Each skill includes:

  • •Context reduction metrics — How much context is saved vs. direct tool calls
  • •Model recommendations — Which models work best for each step
  • •Estimated costs — Approximate provider costs per run
Discover skills using relay_skills_list. The agent can call this tool to see all available patterns before building a workflow.

Available Skills

SkillWithout RelayPlaneWith RelayPlaneReduction
invoice-processor~15k tokens~500 tokens97%
content-pipeline~20k tokens~3k tokens85%
lead-enrichment~10k tokens~1k tokens90%

invoice-processor

Extract, validate, and summarize invoice data using a multi-model AI workflow.

Context Reduction
97%
Est. Cost
$0.02-0.05

Models Used

StepModelWhy
Extractopenai:gpt-5.2Vision-capable, best at structured extraction
Validateanthropic:claude-sonnet-4.5Strong reasoning for math verification
Summarizeopenai:gpt-5-nanoFast and cheap for simple text generation

Usage

1relay_workflow_run({
2  name: "invoice-processor",
3  steps: [
4    {
5      name: "extract",
6      model: "openai:gpt-5.2",
7      prompt: "Extract all invoice fields: number, vendor, date, line items, totals.",
8      schema: {
9        type: "object",
10        properties: {
11          invoiceNumber: { type: "string" },
12          vendor: { type: "string" },
13          date: { type: "string" },
14          lineItems: { type: "array" },
15          total: { type: "number" }
16        }
17      }
18    },
19    {
20      name: "validate",
21      model: "anthropic:claude-sonnet-4.5",
22      depends: ["extract"],
23      prompt: "Verify line item totals match invoice total. Flag any discrepancies."
24    },
25    {
26      name: "summarize",
27      model: "openai:gpt-5-nano",
28      depends: ["validate"],
29      prompt: "Create 2-sentence executive summary for finance approval."
30    }
31  ],
32  input: { fileUrl: "https://example.com/invoice.pdf" }
33})

content-pipeline

Research, draft, and refine content with a multi-stage pipeline.

Context Reduction
85%
Est. Cost
$0.05-0.15

Models Used

StepModelWhy
Researchperplexity:sonar-proReal-time web search for current info
Draftanthropic:claude-sonnet-4.5Strong writing capabilities
Refineanthropic:claude-sonnet-4.5Polish and improve draft

Usage

1relay_workflow_run({
2  name: "content-pipeline",
3  steps: [
4    {
5      name: "research",
6      model: "perplexity:sonar-pro",
7      prompt: "Research key points about: {{input.topic}}. Include facts, statistics, and trends."
8    },
9    {
10      name: "draft",
11      model: "anthropic:claude-sonnet-4.5",
12      depends: ["research"],
13      prompt: "Write a {{input.format}} about {{input.topic}} using: {{steps.research.output}}"
14    },
15    {
16      name: "refine",
17      model: "anthropic:claude-sonnet-4.5",
18      depends: ["draft"],
19      prompt: "Improve clarity, fix issues, add engaging intro: {{steps.draft.output}}"
20    }
21  ],
22  input: { topic: "AI in healthcare", format: "blog post" }
23})

lead-enrichment

Enrich and score leads using AI analysis combined with external data sources.

Context Reduction
90%
Est. Cost
$0.02-0.04

Models Used

StepModelWhy
Extractopenai:gpt-5-nanoFast extraction of email domain
LookupMCP toolCRM or data provider lookup
Scoreanthropic:claude-sonnet-4.5Intelligent lead scoring

Usage

1relay_workflow_run({
2  name: "lead-enrichment",
3  steps: [
4    {
5      name: "extract",
6      model: "openai:gpt-5-nano",
7      prompt: "Extract the company domain from email: {{input.email}}"
8    },
9    {
10      name: "lookup",
11      mcp: "clearbit:companyLookup",
12      params: { domain: "{{steps.extract.output}}" },
13      depends: ["extract"]
14    },
15    {
16      name: "score",
17      model: "anthropic:claude-sonnet-4.5",
18      depends: ["lookup"],
19      prompt: "Score this lead 1-100 based on company data: {{steps.lookup.output}}"
20    }
21  ],
22  input: { email: "john@acme.com" }
23})

Coming Soon

Additional skills in development:

  • •document-qa — RAG-based document Q&A (75% context reduction)
  • •code-review — Multi-stage code review (80% context reduction)
  • •email-classifier — Classify and route emails (95% context reduction)

Next Steps