How to install 10 Hermes agent skills for cheaper runs

TL;DR

Install three starter skills tonight to cut your agent running costs and unlock new abilities. Defuddle cleans web pages before reading, Caveman reduces output tokens, and Codebase Memory keeps heavy repos indexed so your agent queries a map instead of re-reading files. These three changes dramatically improve efficiency and scale for ongoing sessions.

Transcript

I just made this in 1 minute using a single text file. Now my agent never forgets anything I ever tell it. And I have 10 of these running on my agent right now. Three of them cut my agent's costs down, five gave it abilities it did not have at all before, and the last two turned it from one chatbot into an entire team. So, in this video, I'm going ... Read More

Key Insights

• Defuddle is a starter skill that strips pages to a clean reader mode markdown to save tokens.
• Caveman is an ultra compressed output mode that reduces output tokens by up to 65 percent, lowering costs over time.
• Codebase Memory indexes a whole repo once and then queries the map instead of re-reading files, cutting token usage dramatically.
• Agent Reach fixes access to external sites like X, Reddit, and GitHub by providing backchannel access with zero API fees.
• Marketing Skills bundles SEO, copywriting, and conversion knowledge to guide choices rather than just producing writing.
• Open Montage analyzes video structure and provides a cost estimate and plan before spending, aligning outputs with user goals.
• Humanizer rewrites AI-produced text in a natural voice to avoid obvious AI drafting cues, used sparingly and strategically.
• Superpowers acts as an operating system for agents by offering a framework and a method to decompose work for distributed execution.

Questions & Answers

Q: What is Defuddle and how does it help reduce costs?

Defuddle is a tool that strips a webpage down to a clean reader mode markdown before your agent reads it. This prevents the agent from processing unnecessary navigation, ads, and banners, which would otherwise increase token usage across turns. By reducing input content, the agent spends fewer tokens on initial reads, lowering overall costs from the first page read.

Q: How does Caveman impact output costs and why is that important?

Caveman provides an ultra compressed output mode, cutting output tokens by a significant margin. Since output costs are higher than input costs and are included in every subsequent turn, reducing the size of the response directly lowers ongoing expenses. The video demonstrates that the right compression can preserve information while reducing token usage.

Q: What does Codebase Memory do for large repositories?

Codebase Memory creates a memory graph of a large codebase so the agent can query the map rather than repeatedly reading files. This prevents repeated, expensive reads and makes exploration across many languages more token efficient. The result is substantial token savings over sessions with big repos.

Q: Why is Agent Reach important for research tasks?

Agent Reach addresses the limitation that a local agent cannot access certain services like X, Reddit, or GitHub from cloud environments due to API restrictions. It provides a failover path and backends to access these platforms, enabling the agent to research competitors and gather information that would otherwise be blocked, increasing capability without incurring extra API costs.

Q: What value does Marketing Skills bring to an AI agent?

Marketing Skills bundles comprehensive judgment areas such as SEO, paid ads, conversion optimization, and copywriting. This set helps the agent understand what makes a landing page effective or a subject line successful, guiding decision making beyond generic writing to produce outcomes that improve performance in real-world campaigns.

Q: How does Open Montage contribute to cost control and project planning?

Open Montage analyzes a video’s transcript, pacing, structure, and tone, then suggests two to three concepts with a tool path, cost estimate, and preview before any asset is spent. This pre-approval step helps avoid wasted resources by aligning the agent’s output with a clear, vetted plan.

Q: What is Humanizer and when should it be used?

Humanizer rewrites AI-generated text to sound more natural by applying an authentic human voice and updating when sources change. It acts as a quality gate that can help you publish content without obvious AI cues, but it should be invoked late in a session to avoid cluttering the conversation with long rewrites early on.

Q: What makes Superpowers a core component of an agent framework?

Superpowers is described as a skills framework with an actual methodology for breaking work down and executing it. It functions like an operating system for the agent, providing structured processes for planning, delegating subtasks, and coordinating multiple agents or workers to execute complex tasks efficiently.

Summary & Key Takeaways

• The video shows a practical method to reduce agent costs by using three starter skills that dramatically cut both input and output tokens while enabling new capabilities.

• The presenter argues that adding a set of targeted skills can transform a single chatbot into a team and a broader toolset, with ongoing cost savings.

• The overall message is that skills are free to install, but the value comes from choosing and using the right ones to avoid clutter and wasted resources.