Week 3 of 6
Agentic Creativity
Skills, MCPs, and Visual Capabilities
Live on Zoom · 2.00–4.30pm ET
30 August 2026
Derrick Schultz
Course schedule
Week 1 · Aug 16
Introduction to agentic creativity and Hermes Agent setup
Week 2 · Aug 23
Models, text generation, and everyday tasks
Week 3 · Aug 30
Skills, MCPs, and visual capabilities
Sep 6
No class
Week 4 · Sep 13
Long-running agents and how they change over time
Week 5 · Sep 20
An introduction to agentic coding
Week 6 · Sep 27
Show and tell
Today we give the agent two things it did not have last week: reusable methods and new ways to work with images and video.
Today’s schedule
2.00
Review, questions, and a look at the group agent
2.20
Skills: reusable instructions for an agent
2.45
MCPs: connecting an agent to live tools and services
3.05
Break
3.10
TITLES lab: create images, edit images, and make video
3.50
Vision and visual creation for our class agent
4.15
Work time, homework, and questions
today’s question
How do you give an agent a way of working, not just another prompt?
A skill can hold a repeatable method. An MCP can connect that method to a live capability. Together they let an agent do more than talk about a task.
01
Skills
2.20–2.45 · 25 minutes
a useful definition
A skill is a reusable set of instructions, references, and checks for a kind of work.
It gives the agent a tested approach without putting every detail into every conversation.
the problem skills solve
Good prompts disappear. Good methods should not.
When you discover a reliable process—how you organize research, prepare a pitch, or make a contact sheet—you can turn it into a skill.
Next time, the agent can load the instructions only when that task comes up. You keep the method, not just one successful chat.
Make a contact sheet of these images, label each one, and keep the originals untouched.
SKILL
When asked for a contact sheet:
1. inspect the folder
2. preserve originals
3. make labeled output
4. verify every file appears
The skill captures a sequence, constraints, and a definition of done.
a skill is more than a prompt
Useful skills make the important choices explicit.
01
When to use it
The trigger: what kind of request calls for this method?
02
Procedure
The ordered moves that make the work repeatable.
03
References
Details, examples, templates, or source material loaded only when needed.
04
Verification
How the agent checks that the output is usable before it says it is done.
Skills use progressive disclosure: the agent sees a small index first, then opens the full instructions only for relevant work. · Hermes: Skills system
inside SKILL.md
The frontmatter helps Hermes find the skill. The body teaches Hermes how to work.
Frontmatter names and describes the skill. Optional metadata can identify its version, platforms, tags, required tools, or configuration.
Markdown sections explain when the skill applies, the procedure to follow, known failure modes, and how to verify the result.
SKILL.md is the required file. A skill directory can also carry optional references/, scripts/, templates/, and assets/. · Hermes: creating skills
---
name: visual-critique
description: Ground feedback in visible evidence
---
# Visual Critique
## When to Use
Use when someone asks for feedback on an image.
## Procedure
1. Describe visible evidence.
2. Compare it with the stated intention.
## Pitfalls
Never invent details that are not visible.
## Verification
Cite visual evidence for every suggestion.
Hermes begins with a broad catalog of bundled skills.
Research
Methods for finding and retrieving material, such as arXiv research and structured web work.
Creative work
Task-specific methods for making things, from ASCII art and diagrams to documents and media workflows.
Projects & code
Procedures for planning, working with repositories, using coding agents, and managing pull requests.
Agent operations
Skills for configuring Hermes itself, computer use, connected tools, and multi-agent workflows.
A fresh Hermes profile is seeded with the bundled catalog; optional skills are separate and must be installed. Use /skills or hermes skills list to see what your installation has. · Hermes: bundled skills catalog
where skills live
Skills in Hermes Desktop
Open Capabilities → Skills to see what is installed, inspect its instructions, and enable or disable it.
The Skills Hub appears below your installed list. Use it to search bundled, optional, and community sources before adding a skill to your agent.
a reusable skill from GitHub
Inspect a skill before you install it.
Install one skill
hermes skills inspect
owner/repo/skills/my-workflow
hermes skills install
owner/repo/skills/my-workflow --now
Use a repository repeatedly
hermes skills tap add owner/repo
hermes skills search my-workflow
--source owner/repo
hermes skills install
owner/repo/my-workflow --now
--now activates it in the current session; otherwise begin a new session or use /reset. Read the instructions and review the scan before installing third-party skills. · Hermes: Skills Hub and GitHub sources
the command line is optional
You can ask Hermes to inspect and install the skill for you.
Give it the complete skill identifier or URL. Ask it to explain what the skill does, show you the security scan, and pause before installing.
The agent performs the same underlying workflow—you are delegating the commands, not the judgment.
You can also type /skills install <identifier> --now directly in chat. · Hermes: working with skills
Inspect owner/repo/skills/my-workflow. Tell me what it does and flag risky instructions. If the scan is clean, ask before installing it for this session.
I’ll inspect the source and scan result first.
Hermes skills · inspect
It contains a focused workflow and no dangerous findings. Install it with --now?
Yes.
Hermes skills · install
example · visual review
A skill can teach the agent how to look at creative work.
When to use
Use when someone wants feedback on an image, sequence, or edit.
Procedure
1. Describe what is visibly present before interpreting it.
2. Name the strongest formal choices: framing, light, colour, rhythm.
3. Compare the result with the stated intention.
4. Offer 2–3 concrete next experiments.
Check
Do not invent details that are not visible.
The skill gives an agent a critique posture. A vision-capable model gives it the ability to inspect the actual image.
procedural memory
Hermes can turn a hard-won workflow into a skill for next time.
When it works out a non-trivial sequence, resolves a dead end, or you correct its approach, Hermes can use skill_manage to create or update the instructions.
That is different from remembering a fact. A skill preserves how to work; short-lived or durable facts belong in memory.
We finally got the grant packet right. Save that process.
I’ll turn the successful workflow into a focused skill: inputs, order, checks, and the mistakes we avoided.
skill_manage · create or patch a skill
Next time, I can load that method when the task appears.
This is not automatic for every chat: a skill should capture a reusable method. By default Hermes can write skills; enable skills.write_approval when you want to review each change first. · Hermes: agent-managed skills
Put information where it will be useful.
A prompt
One request, one moment. Use it for the subject, constraints, and outcome of this particular task.
AGENTS.md
Durable rules for everything in one project: structure, commands, tone, and limits.
A skill
A repeatable method that only matters for a particular kind of work.
An MCP
Live access to a service or tool that performs actions or retrieves current information.
five-minute exercise
Find one process you repeat and write its first skill.
Choose a real recurring task. It can be creative, administrative, or research-based.
Write four headings: When to use, Procedure, Pitfalls, and Verification.
Do not try to make it universal. A useful first skill is narrow enough to test next week.
Pair up if you want: explain the task to someone else and notice the steps you forgot to say out loud.
02
MCPs
2.45–3.05 · 20 minutes
model context protocol
An MCP is a bridge between an agent and a live service.
It describes the tools a service offers, the inputs they accept, and the results they return. The agent can then decide when to use them.
what happens in practice
The model plans. The MCP makes the specific action possible.
You ask for an outcome. The agent recognizes a useful capability, sends a structured request to that tool, reads the result, and continues.
The MCP is not the model. It is the connection that gives the model a specific way to act in the world.
Make a 16:9 image of a night garden in a painterly artist model.
I’ll find a compatible model, then generate a draft.
TITLES MCP · find models
TITLES MCP · generate image
Here are four directions. Want to develop one or try another model?
connect one service deliberately
Install the MCP, choose only the tools you need, then use the auth strategy its service requires.
A safe first install
hermes mcp catalog
hermes mcp install <name>
Choose only the needed tools.
hermes mcp test <name>
/reload-mcp
Authentication choices
API key / header
For a service account or model API.
OAuth
For a hosted service acting as you; approve in a browser.
Local stdio environment
Pass only the exact variables a local server needs.
Let Hermes prompt for secrets; do not put keys in chat, slides, or source control. OAuth tokens are stored locally and refreshed by Hermes. Use a tool allowlist and start with a read-only test. · Hermes: MCP setup and authentication
a larger visual tool shelf
The fal.ai MCP gives an agent access to a live catalogue of image, video, audio, and other media models.
First, find the right tool
search_models
Find a model by task or medium.
get_model_schema
See the inputs and outputs.
get_pricing
Check cost before running it.
Then, make and retrieve
upload_file → run_model
Send a source and return a result.
submit_job → check_job
Start a longer video job, then retrieve it.
The hosted server is mcp.fal.ai/mcp. It uses your own API key, and model runs are charged at their listed cost. · fal.ai: MCP documentation
Skills decide how to work. MCPs supply the capability.
A skill says
Before a visual generation, identify the desired mood, format, audience, and whether the result needs to move.
An MCP says
Here are the tools for finding models, generating an image, editing it, or making a video.
The agent does
Choose a tool, send the right inputs, inspect the result, and continue the process with you.
You remain responsible for
The intention, the references you share, the spend you approve, and the final selection.
a safety habit
Every MCP is also a new permission.
Before connecting
What account can it reach?
Can it read private material?
Can it write, publish, or spend?
Where does the key live?
What will the service retain?
While using it
Read the tool description.
Check the inputs.
Start with a small test.
Notice confirmations and cost.
Review the output.
Do not paste passwords or API keys into a chat. Use the service’s intended setup and keep early experiments low-stakes. · Hermes: MCP
03
TITLES lab
3.10–3.50 · 40 minutes
the model changes the image
With artist models, choosing a model is part of directing the work.
The same words can produce very different images with different models. Start with the visual language, not just the subject.
choose the right capability
Image and video models solve different parts of a visual process.
Image
Generate a still
Best for exploring composition, style, objects, and a single decisive frame.
Edit
Revise a source
Keep some visual information while changing a subject, material, or setting.
Video
Animate an image
Turn one strong frame into a short moving moment with a clear motion direction.
Vision
Read an image
Describe, compare, and critique what is visible so the next prompt can be more precise.
These are complementary capabilities, not a single ladder from worse to better.
before the lab · one-time setup
Ask Hermes to connect the production TITLES MCP with OAuth.
Give Hermes the complete server address and name the authentication method. It can perform the setup while you keep control of account authorization.
Hermes opens the TITLES sign-in page. Sign in, review the access request, approve it, then return to Hermes and select the tools.
If you prefer Terminal: hermes mcp add titles --url https://titles.xyz/mcp --auth oauth. If authorization expires, run hermes mcp reauth titles. Never share the callback URL; it contains a temporary code. · Hermes: OAuth-authenticated MCP servers
Install https://titles.xyz/mcp as an MCP server named titles. Use OAuth and ask before enabling the tools.
I’ll add the server and open TITLES authorization in your browser.
Hermes MCP · add titles · OAuth
Approve access in the browser, then return here to choose which TITLES tools to enable.
Approved. Enable the tools we need for today’s image and video lab.
TITLES MCP · connected
verify with a harmless read
Check the account, not just the connection.
A successful server test only proves that Hermes can reach TITLES. Ask a question that confirms which account OAuth authorized.
If the handle is missing or wrong, open the TITLES MCP settings and authorize again before generating or publishing anything.
This read-only check catches the most important setup mistake: connecting the right service through the wrong account.
Who am I on TITLES?
I’ll check the account connected through the TITLES MCP.
titles_get_me · called
You’re signed in as @yourhandle.
Great. What image and video tools are available to us?
TITLES MCP · list enabled tools
before the lab · add the methods
Install the TITLES skills so Hermes knows how to turn the tools into workflows.
The MCP provides live actions such as searching models and generating media. The skills repository provides reusable methods for choosing and sequencing those actions.
Inspect the repository and its scan before installing. For today, make sure generate-image and animate-image are available.
github.com/titlesnyc/titles-skills · Ask Hermes to activate the skills now, or begin a new session after installation.
Inspect https://github.com/titlesnyc/titles-skills. Show me the security scan, then ask before installing the skills.
I’ll inspect the repository and summarize what the skills can do.
Hermes skills · inspect repository
The scan is clear. Install the TITLES skills and activate them for this session?
Yes. We need generate-image and animate-image for the lab.
Hermes skills · install and activate
skills + MCPs in one run
One request can coordinate a method and several live tools.
The skill supplies the sequence and the criteria for choosing. The TITLES MCP supplies catalogue search, image generation, and video generation.
You can interrupt between steps, change the selection, or stop before a tool that spends more.
The agent is coordinating the run; it is not hiding the tools. Watch the calls and keep the creative decisions visible.
Find a painterly folk-horror model, make four stills, choose the strongest composition, then animate it with one slow push forward.
visual concept skill · plan and selection criteria
TITLES MCP · search models → generate four images
The second still has the clearest silhouette and strongest depth. Animate that one?
Yes—keep the figure still and move only the camera and mist.
TITLES MCP · generate video
make a strong still first
Give the agent direction it can turn into a useful first generation.
Describe the subject, but also name the visual language: material, palette, spatial arrangement, light, and intended format.
Then ask for several distinct directions. A small set of deliberate variations is more useful than an endless pile of near-duplicates.
Subject
a public bench in a night garden
Visual language
paper-cut shadows, cobalt and orange,
humid air, theatrical pool of light
Format
16:9 image for a projected opening slide
Variation request
Give me four directions with clearly different
compositions, not small prompt variations.
image editing
Editing is useful when you want change without starting from nothing.
Tell the agent what must remain recognizable and what can change. The stronger the anchor, the more specific the requested change should be.
image-to-video
A moving image needs a motion decision, not just a beautiful frame.
Start with a still you would be happy to hold on screen. Then choose one main movement: camera motion, environmental motion, or character motion.
Keep the direction simple. Several competing movements often make a short clip feel unstable rather than alive.
Prompt: slow push forward; red threads and small bells move in a light wind; mist drifts while the figure remains still. Choose the motion before you ask for the video.
a quality-to-cost video pick
H3 Max is a strong first choice for short, high-quality video when you want to iterate.
fal’s post-trained MiniMax H3 variant is tuned for prompt adherence and aesthetics, with text-to-video and image-to-video inputs. It makes 480p or 768p clips with native audio.
Its regular 768p list price is $0.08 per second: about $0.40 for a five-second experiment. That makes it practical to test a few clear motion directions before committing to a more demanding production.
Best for
Fast image-to-video and text-to-video studies
Output
480p or 768p · native audio · up to 15 seconds
Speed
A five-second clip in under three seconds, according to fal
Use another model when
You need 2K, reference-to-video, or video editing.
The price is a regular listed rate, not the temporary launch discount. Confirm current pricing in the tool before generating. · fal: H3 Max details and pricing
lab · work in pairs if helpful
Use the TITLES MCP to make one short visual experiment.
1. Pick a model for a visual language. Ask the agent to explain why it fits your brief.
2. Generate a still. Make a few distinct directions, then select one worth developing.
3. Choose a next move. Edit it, restyle it, or give it a single motion idea for video.
4. Keep the receipt. Save the prompt, model, output, and what you would try next.
Start small. Watch the tool calls and read any cost confirmation before you approve it.
Give the agent decisions it can work with.
Intention
What should this image or clip make someone notice, feel, or understand?
Visual language
Name a model, reference, material, palette, composition, or level of abstraction.
Boundaries
Say what must stay, what cannot appear, and what format or duration the work needs.
Next experiment
Ask for a meaningful variation: change the frame, point of view, time, palette, or motion—not everything at once.
04
Our class agent learns to see and make
3.50–4.15 · 25 minutes
a new kind of collaborator
Vision lets an agent respond to the work in front of it. Creation lets it propose a next image or moving study.
The important part is the loop: look closely, make a specific observation, try one visual move, then look again together.
two independent model slots
The model that sees does not have to be the model that speaks.
Your main model handles the conversation, planning, tool loop, and final response.
Hermes can send images and browser screenshots to a separate auxiliary vision model, then return those observations to the main agent.
For this class, use Gemini 3.7 Flash for vision. You can keep whichever main model you prefer for agent responses. Model identifiers vary by provider. · Hermes: auxiliary models
MAIN AGENT
your preferred response model
conversation · planning · tool use
AUXILIARY → VISION
Gemini 3.7 Flash
images · screenshots · visual inspection
SET IT
hermes model
→ Configure auxiliary models
→ vision
→ choose provider + Gemini 3.7 Flash
how we will use vision
Ask for observation before interpretation or advice.
When we give the group agent an image, it should first describe what is visibly there: framing, objects, colour, light, type, density, and movement.
Only then should it connect those observations to your stated intent and offer a next experiment. That keeps the critique grounded in the actual work.
Here is my poster draft. Does the title compete with the image?
vision · reads composition and text placement
The title occupies the brightest upper-left area, where the face also has its strongest contrast. Try moving the title into the darker lower band, or reduce its weight before changing the image.
Make a second version with that lower-band title treatment.
TITLES MCP · visual study
we are adding a small capability stack
Each addition has a distinct role in the process.
Skill
Visual critique
A shared way to describe work and suggest experiments without inventing what is not there.
Vision
Image input
The ability to inspect an image, compare versions, and ground feedback in visible evidence.
MCP
TITLES tools
Live access to artist models for image generation, editing, and video creation.
Memory
Project thread
A brief record of the intention, selected work, and next question for each participant.
a shared working agreement
Our agent should make the process more legible, not take it over.
1. We say what we are trying to test.
2. The agent describes visible evidence before judging it.
3. It offers a small number of specific next experiments.
4. We choose whether to make one, revise the prompt, or stop.
5. We save the model, prompt, output, and decision.
THE AGENT DOES NOT DECIDE WHAT IS FINISHED.
This is a studio critique tool and a sketching collaborator. The intention and final judgment stay with the artist.
05
Take it forward
4.15–4.30 · 15 minutes
For next week
Finish one skill
Make the small skill you started today runnable. Test it once and revise the part that was unclear.
Keep one visual receipt
Bring a TITLES image or video experiment with its model, prompt, output, and one sentence about the next iteration.
Give feedback with evidence
Use the visual critique approach on your own work or someone else’s: describe first, interpret second, propose one concrete move.
Notice a long-running task
What would your agent need to remember, check, or revisit over several weeks? We will use those examples in Week 4.
Keep learning
Skills
Hermes Skills System — on-demand knowledge, progressive disclosure, and skill structure
MCPs
Hermes MCP guide — how Hermes discovers and connects external servers
TITLES
TITLES MCP — artist models and the visual tools used in today’s lab
This deck
that’s week three
Make a method. Connect a capability. Keep your judgment.
next week
Long-running agents and how they change over time
13 September · 2.00pm ET
Questions in Discord, or derrick@titles.xyz · artificial-images.com