Three Hermes AI Skills: Zoltar, Antiscammer, Book-Mirror

Three user-created Hermes AI skills are available in the jaldps/hermes-skills repository: Zoltar, Antiscammer and Book-Mirror. Zoltar generates astrological and divination reports, Antiscammer automates responses to mass-market scams on messaging apps, and Book-Mirror produces two-column chapter analyses linked to a user’s stored context.

Zoltar uses established astronomy and occult libraries. Given a birth date, time and city, it runs Kerykeion and the Swiss Ephemeris to compute planetary positions by sign and degree, house placements with the Placidus system, an aspect grid, element and modality breakdowns and retrograde flags. The skill can produce interpretations that reference the specific placements. Zoltar also offers live daily horoscopes via the Ohmanda API, structured tarot draws using a 78-card Rider-Waite-Smith dataset, an I-Ching coin-casting implementation that returns named hexagrams with Judgment and Image texts, Pythagorean numerology charts and Elder Futhark rune casting with rules for reversed positions. Users can request a styled HTML report and save it to a desktop.

Antiscammer uses Hermes’s browser control to send automated replies on WhatsApp Web, Telegram Web and Discord. One mode can transmit the full Shrek screenplay, 3,679 lines, one message at a time; an alternative uses the Bee Movie script. Another mode generates long, deliberately bureaucratic email replies written in an archaic solicitor style. The skill is intended for mass-market scams such as inheritance, fake bank and crypto recovery messages. The designers advise against using it on sophisticated phishing attempts or any interaction that could expose live credentials.

Book-Mirror creates a two-column chapter-by-chapter analysis. The left column records the author’s points, quotes and frameworks; the right column maps those points to the user’s real life using names, roles, projects and prior session excerpts drawn from Hermes’s memory. If memory is limited, the agent asks targeted questions to fill gaps. The skill extracts text from PDFs and EPUBs with PyMuPDF and ebooklib, processes chapters sequentially, and can fetch public-domain texts when given only a title. For longer books the skill processes chapters in order and may take more time to complete.

All three skills are stored in the jaldps/hermes-skills collection and install by placing a SKILL.md file in ~/.hermes/skills/ or by asking Hermes to fetch the repository URL. At the time the skills were noted the Hermes community had gathered roughly 134,000 stars across GitHub repositories, and the repository catalog includes tools ranging from developer workflows to games.

Developers and users should consider legal and ethical limits before deploying automated replies or targeted personal analyses. Antiscammer’s documentation cautions against using the tool for phishing that targets live accounts or involves credential disclosure. Book-Mirror’s output depends on the depth of the agent’s stored context and on user consent to share personal details.