👤 For pipe admins 🔐 Available on all plans 🎯 If you want to leverage the power of artificial intelligence to enhance your processes with automations. |
Note: If you’d like to learn more about what AI Agents are, click here.
Overview
What are Skills?
Skills (or AI Skills) are intelligent components that extend the capabilities of AI Agents in Pipefy. They act as knowledge sources or ways for the agent to interact with Pipefy, enabling it to make more accurate decisions and execute tasks with richer context.
Why are they important?
By using one or more skills, an AI Agent can query external data beyond the current card—such as attachments or records from other pipes—to perform its tasks more precisely.
This enhances agent autonomy and reduces reliance on manual user actions.
🔍 Example: An AI Agent that needs to respond to a question may use a skill to consult a PDF document or search records in another Pipe before replying.
Available Skills
📄 Skill: Access to documents
Allows the AI Agent to consult a specific document and extract relevant information that helps execute its task.
Parameters:
Name: Provide a clear, descriptive name (e.g., “Consult Product Manual”).
Description: Explain in what scenario the Agent should use this skill (e.g., “Use when needing to answer technical product questions”).
Document Attachment: Upload the PDF file.
📌 Recommendation: Choose short or well‑structured documents. The more concise the content, the better the Agent’s performance.
📊 Skill: Access to pipe data
Allows the Agent to search for information in pipes and use that data as a reference for its task.
Parameters:
Name: A clear and descriptive name (e.g., “Search purchase history”).
Description: When the Agent should trigger this search (e.g., “Use to recommend products based on previous purchases”).
Source Pipe: Choose the pipe where data will be retrieved—from another pipe or the Agent’s own pipe. Select which pipe to query.
Search Filters: Define conditions to filter relevant cards. Examples:
Current phase = “Concluded”
Field “Customer” = matches the dynamic value from the instructions
That value will be filtered using a dynamic field inserted into the prompt for each card where the AI automation runs.
Note: If you use more than one filter, the logic is always “AND”.
Returned Information: Choose which card fields should be used by the Agent (e.g., product name, purchase date, amount).
📌 Recommendation: Be specific with filters and fields to ensure the Agent receives precise and useful data.
⚠️ Important limitations
The pipe data access skill can return a maximum of 30 cards per execution.
If more than 30 cards match the configured filters, the 30 most recently created cards will be returned.
How Skills integrate with AI Agent behavior
Skills are automatically triggered by the AI Agent based on the task instructions it receives and the conditions configured for each skill. The Agent assesses when additional information is needed and decides which skill to apply based on the automation’s context.
This means admins don’t need to program when skills are invoked. You only need to define clear parameters (name, description, filters, etc.), and the Agent will decide when to use them.
Best practices for creating Skills
Use clear names and descriptions: The Agent chooses which skill to use based on them.
Avoid long or generic documents. Split content across multiple skills if needed.
Apply precise filters when accessing pipe data: This avoids unnecessary returns and improves performance.
Regularly update referenced documents and pipes to ensure the Agent works with up‑to‑date information.
Use Cases
Example 1: Automated support based on company policy
Task: The Agent responds to questions about the refund policy.
Skill used: Document Access with the PDF of the policy.
Example 2: Analyzing customer interaction history
Task: The Agent recommends an action based on previous interactions.
Skill used: Pipe Data Access, querying cards in a previous support pipe.
Related Content
🎓 Learn more with free courses from Pipefy Academy! |