9 Hermes Agent Skills You NEED to Install Today

Summary of the video (technological concepts + product/features)

The video introduces “9 Hermes Agent skills you need to install today” and explains how installing them can improve an agent’s cost efficiency, capabilities, and ability to act like a coordinated “team” to deliver real business results. It also claims to show installation steps for each tool/skill.

---

1) Photo — web-page content filtering to reduce token usage

• Goal: reduce wasted token consumption during web research.
• Problem: when an agent browses sites, it often ingests lots of irrelevant HTML UI content, such as:
  • headers/footers
  • cookie banners
  • navigation menus
  • ads, popups, and similar clutter
• Approach: parse the HTML and strip irrelevant sections, sending only the meaningful content to the model.
• Benefit: reduces how much unnecessary text is stored/re-read in the context window across multi-turn follow-ups.

---

2) Code Base Memory MCP — fast code understanding via indexing

• Type: presented as an MCP tool (more than a simple “skill”).
• Function: indexes and breaks a GitHub/codebase into queryable parts so the agent can answer feature questions without rereading the entire repo.
• Claims:
  • indexing happens in milliseconds (example given: ~875K files indexed in ~3 minutes)
  • supports multiple languages (e.g., Python, TypeScript/JavaScript, etc.)
  • claims large token savings (stated as ~120X fewer tokens), attributed to retrieval/indexing vs. repeated ingestion
• Install approach: “super simple”—provide the MCP URL to Hermes (the agent) and ask it to read/install.

---

3) Caveman — concise response style to cut output tokens

• Type: a GitHub “skill” designed to make the model respond more concisely.
• Viral claim: ~100,000 stars.
• Token/cost claim:
  • the repo suggests ~65% fewer tokens
  • the video cautions that savings apply mostly to output tokens; input/reasoning tokens may remain similar
• Reviewer stance: not personally tested by the narrator, but commenters/research suggest it may help.

---

4) Humanizer — rewrite text to sound more human

• Purpose: rewrite AI-generated text to be less robotic and more human-sounding.
• Description: includes 25 patterns/logic (details in the repo; not fully reviewed in the video).
• Example observations:
  • removes “telltale” formatting markers like end dashes
  • changes phrasing/structure compared to original AI output
• Usage guidance: not “lightweight,” so use selectively (e.g., social posts, blog posts, content meant for humans).

---

5) Agent Reach — bypass access restrictions like 403

• Motivation: sometimes browsing/retrieval fails due to 403 or “no access for robots” restrictions.
• What it does: helps the agent work around access blocks, retrieve needed information, and answer the user’s query.
• Emphasis: enables research when normal browsing is restricted.

---

6) Marketing skills package — 49 marketing/SEO/conversion-focused skills

• What it is: a “package” of 49 skills aimed at marketing and conversion-oriented content.
• Focus areas listed:
  • marketing, SEO, paid ads
  • buyer psychology
  • cold outreach
  • copywriting
  • content strategy
• Claimed value: improves judgment—helping the agent decide what will make a page/content convert versus not.

---

7) GA Code / Gatekeeper code — sub-agent coding workflow under Hermes control

• Description: an extension for Hermes.
• Goal: allow Hermes to trigger a more advanced “coding agent” environment for system development, beyond Hermes’s typical day-to-day automation role.
• Workflow claim (sub-agent lifecycle):
  • plan → execute → review → fix → deliver
• Setup notes:
  • integrates via logging with a provider if the user has a subscription
  • claims it doesn’t require extra APIs/connections

---

8) Superpowers — software engineering framework/methodology

• What it is: a GitHub framework for agent-driven software development with process controls (planning, review, verification).
• Core idea: follow a methodology rather than “start building immediately.”
• Mentions: includes example install commands across different tools/environments (e.g., Cursor and others).

---

9) Hermes Agent plugin — Git worktrees + agent-driven dev lifecycle

• Description: includes:
  • brainstorming using Git worktrees
  • writing plans
  • owning an agent-driven development loop
  • executing plans, testing, reviewing, and delivering
• Framing: positions this as a professionalized development workflow.

---

Bonus / last item (not a “skill”): platform aggregating many APIs/skills/models

• What it is: a platform providing 5,000+ APIs/skills/AI models.
• Claimed impact: transforms Hermes into a more capable “team” agent with premium tools and information access it couldn’t previously use.

---

Main speakers/sources (as indicated in the subtitles)

• Primary narrator/host: the person speaking throughout the video.
• GitHub authors/creators mentioned:
  • Julius (creator of Caveman)
  • GitHub maintainers/teams for each named tool (e.g., Code Base Memory MCP, Agent Reach, Gatekeeper code/GA Code, Superpowers), though no individual names are provided beyond Julius.