This document describes the agents and skills available within the footnotes project (AI Footnotes).
The extension embeds a terminal linked to a local pseudo-terminal (PTY) running the Gemini CLI. The interactions are orchestrated by an agent that leverages specialized "skills" to perform tasks directly on the user's local machine.
Skills are modular sets of instructions and tools housed in the skills/ directory.
- Location:
skills/librarian/ - Description: Resolves Decentralized Identifiers (DIDs) and searches for information using ODRL services. It provides semantic search capabilities and seamless integration with metadata repositories.
- Location:
skills/notetaker/ - Description: Saves highlighted or provided text as a note in a local directory. Clicking "Save Note via Gemini" in the extension UI triggers this skill, creating uniquely named notes for later reference.
- Location:
skills/interpretator/ - Description: Finds contradicting interpretations of a given fragment of text, comparing the pros and cons of each perspective to provide a balanced analysis.
- Location:
skills/transcripter/ - Description: Extracts full English transcripts from YouTube videos using their video ID. This allows LLMs to summarize or analyze video content directly from YouTube links without hallucinating.
To add a new skill to the footnotes ecosystem:
- Create a new subfolder in
skills/(e.g.,skills/my_new_skill/). - Add a
SKILL.mdfile within the folder containing the specific instructions and YAML frontmatter for the skill. - Place any helper scripts in a
scripts/subdirectory within your skill folder. - Keep all instructions modular and follow the conventions described in existing skills.