Overview

The freeact-skills project provides a curated set of predefined skill modules for the freeact agent system. These modules must be pre-installed on ipybox to be accessible by freeact agents.

Skills

Currently in its early development phase, freeact-skills offers a set of core skills and specialized skills. To use a skill in a freeact agent application, reference it by its module name (see freeact tutorials for details). For example, the zotero skill can be referenced as freeact_skills.zotero.api.

Skill module sources are retrieved from ipybox and incorporated into the agent's model context. By convention, freeact-skills packages are split into an api module containing the skill interface, including a detailed description and optional usage examples, and one or more implementation modules containing the underlying logic.

This convention ensures that freeact agents only receive relevant information from the api module while internal implementation details remain hidden. However, it is merely a guideline; you are free to present any Python code as a skill module, depending on your context requirements and the level of detail you want the agent to process.

Core skills

• freeact_skills.search.google.stream.api

  Provides generative Google search via the Gemini 2 API. Search output is streamed, enabling clients to display results as they are generated.

• freeact_skills.search.google.api

  Provides generative Google search via the Gemini 2 API. Search output is returned as a string for direct processing by agents in code actions. Maintained for experimentation with Gemini 2 based freeact agents.

Both Google search skills require a GOOGLE_API_KEY or GEMINI_API_KEY, obtainable from Google AI Studio.

• freeact_skills.search.perplexity.api

  Provides generative Perplexity search via the Perplexity API. Search output is streamed for real-time display and includes indexed source links in results. Requires a PERPLEXITY_API_KEY, obtainable from the Perplexity API Settings page.

Specialized skills

• freeact_skills.zotero.api

  Provides access to a Zotero group library via the Zotero API. Requires a ZOTERO_API_KEY which can be obtained from the Zotero Security Settings page. A ZOTERO_GROUP_ID environment variable with a group library ID must also be defined. Here's what a freeact agent says about this skill module when asked to describe it:

  The Zotero module provides a way to interact with Zotero group libraries programmatically. Here's what it can do in simple terms:

  1. Basic Library Access

     • Connect to a Zotero group library using an API key
     • Sync local data with the latest changes from Zotero
     • Browse through collections (folders) and documents

  2. Document Management

     • Access document details like titles, URLs, abstracts, tags, and publication dates
     • View documents organized in collections and sub-collections
     • Find specific documents using their unique keys
     • Add web links to documents with optional titles and notes

  3. Collection Organization

     • Navigate through collection hierarchies (folders within folders)
     • List all documents in a collection, including those in sub-collections
     • Keep track of versions for collections and documents
     • Access the full library structure starting from the root collection

  4. Attachments

     • Work with two types of attachments:
       • Notes: Text annotations attached to documents
       • Links: Web URLs with optional titles and notes

  The module is primarily designed for reading and organizing academic references in a Zotero group library, with some ability to add new content (like links). It maintains a local copy of the library that can be synchronized with Zotero's servers to get the latest updates.

• freeact_skills.reader.api

  Provides access to Readwise Reader via the Readwise Reader API. Requires a READWISE_API_KEY which can be obtained from the here. Here's what a freeact agent says about this skill module when asked to describe it:

  The reader module provides the following capabilities:

  1. Save documents to Readwise Reader in two ways:

     • By URL: You can save any document by providing its URL using save_document_url(). You can optionally specify:
       • Where to save it (location: new, later, or archive)
       • The type of content (category: article, pdf, tweet, or video)
       • Tags to organize the document
       • A note attached to the document
     • By HTML content: You can save HTML content directly using save_document_html() with similar options as above, plus requiring a title

  2. List and retrieve documents from Readwise Reader using list_documents():

     • Can filter documents by location (new, later, archive, or feed)
     • Can get only documents updated after a specific date/time
     • Returns detailed document information including:
       • Basic metadata: ID, title, author, source, summary
       • URLs: Both Reader's URL and original source URL
       • Categories and location
       • Tags and notes
       • Reading statistics (word count, reading progress)
       • Various timestamps (created, updated, saved, opened)
       • Additional metadata like image URL and parent document ID

  3. The module provides a factory function create_readwise_reader() to instantiate the client:

     • Takes an optional API key
     • If no API key is provided, it looks for it in the READWISE_API_KEY environment variable

  This module essentially allows you to programmatically save documents to Readwise Reader and retrieve information about your saved documents, enabling automation of document management and reading workflows.

System skills

• freeact_skills.editor

  Internal system component (not for direct application use). Enables agents to view, create, edit, insert, and undo file edits, supporting interactive skill module development. Based on Anthropic's original implementation.

Status

freeact-skills is in early development. Skill interfaces and implementations may change frequently, relying on the adaptability of freeact agents to handle such changes gracefully.