Configure LLM import steps and guardrails for Celigo flows.
npx skills add https://github.com/celigo/ai --skill configuring-ai-agents
Please help me install this Agent Skill. Skill: configuring-ai-agents Source: https://github.com/celigo/ai/tree/main/skills/configuring-ai-agents Command: npx skills add https://github.com/celigo/ai --skill configuring-ai-agents
This Skill helps you set up LLM-powered import steps and safety/compliance guardrails so your data pipelines can classify, extract, validate, or generate structured results with predictable output.
Tell your AI coding assistant: "Configure an OpenAI AI agent import that extracts invoice fields into strict JSON, and add a PII guardrail to mask emails and SSNs."
High-intent search queries and answers about installing and using this skill.
Configure AI agent imports by setting the correct adaptorType and provider-specific instruction paths to enforce strict json_schema response formats, allowing LLMs to reliably extract and classify unstructured text into structured JSON fields for downstream mapping.
Add PII masking and content moderation guardrails by configuring guardrail imports within your data pipeline. This applies ai_agent guardrails to detect and mask sensitive information like emails and SSNs, ensuring compliance before records are processed.
Yes, you can use OpenAI and Gemini providers via LiteLLM proxy for LLM-powered imports. Both support classification, extraction, and generation tasks, requiring correct provider-specific model configuration paths and optional BYOK via _connectionId.
AI agents use MCP and Celigo Tools during record processing by enabling optional tool use within the import configuration. This allows the LLM to access web search, image generation, and custom tools to classify, extract, or validate records dynamically.
Structured LLM outputs fail validation when strict json_schema response formats are not correctly applied to the flow pageProcessors[] entry. Configuring the correct adaptorType and response mapping ensures consistent field extraction required for downstream validation.
No, you do not need a custom connection, but you can optionally use Bring Your Own Key (BYOK) by configuring the _connectionId parameter. This allows you to securely route OpenAI and Gemini API calls through your own provider accounts.
ChrstprJohnExtract typed, validated data from LLM responses using provider-specific schema strategies.
sanskarbasnetExtract page titles, descriptions, and links as JSON from URLs.
drbobberAccess ATXP paid APIs for web search, AI media generation, and X/Twitter search.
TanStackConvert chat LLM outputs into typed JSON objects via schema enforcement.
