Three directions for the launch image, shown at chat size. Do not publish anything — wait for my pick.What comes backCandidates in the thread with nothing released, and a clear statement of which channel each was prepared for.
Felo SuperAgent on Grok Bot
Grok Bot
Grok BotThe launch image needs approval. The Bot should stop there.
A Bot with a computer can make things, which is exactly why the boundary matters. Felo SuperAgent runs the image work on a persistent canvas and holds the result until you approve it, so a launch asset never leaves without a decision.
- Core unitsBot · skill · routine
- InputA brief in the thread
- Comes back asFiles, plus the run that made them
- BoundaryWritten into the Bot description
The request that saves the method
Save this as a skill called "Launch image". Show three directions,
wait for my approval, and never publish anything without it.A Grok Bot owns a job, a conversation and a schedule, and a skill is the method it runs. Felo skills are the research half of that loop, saved once and reused by every routine.
- Core units
- Bot · skill · routine
- Skills come from
- A saved run, or Settings → Plugins → Yours
- Runs on
- iPhone, Android, and desktop
Add it on Grok BotLearn it from one launch you approved by hand
The method includes the boundary, so teach both at once: generate, hold, approve, deliver. A saved skill that skips the hold is a skill that posts on your behalf.
- Run
Give one Bot a real job
Create a Bot for launch assets, and write the boundary into its description: it prepares images, and it publishes nothing without approval.
- Save
Ask it to record the method
Run one launch by hand — three directions, your pick, then delivery — and save the process with the hold written in as a required step.
- Enable
Turn the skill on where it is needed
Enable it on the launch Bot only. Enabling an image skill on a general chat agent widens what can be published for no benefit.
- Schedule
Attach a routine, then read the first run
Skip the routine at first. Imagery usually needs a person; attach a schedule only if the asset is genuinely repeatable, and read the first run before letting it deliver anything.
A run that stops where your judgement begins
The interesting part of an unattended run is where it refuses to continue. This one holds at the approval boundary, which is the behaviour that makes an image Bot safe to keep.
- Running
Reading the brief
A launch image for three channels, seen first as a chat thumbnail, and it must not look like the last launch's image.
- Running
Three directions
One built from the product itself, one from the claim the launch makes, and one that leads with the date.
- Blocked
Held for your approval
Two directions are ready. Nothing is published, and the run says which channel each one was prepared for.
- Done
Approved, then delivered
The direction you chose went out through the Bot's own channels; the other two stay on the canvas as rejected options with their reasons.
Two prompts that keep the decision with you
Both ask for the hold explicitly, which is what turns an image generator into something you can leave running.
Use the second direction. Adjust it for the smaller format and hold it again before it goes anywhere.What comes backA revised asset at the right size, held at the same boundary rather than sent on the strength of the previous approval.
From a sentence to an approved asset, with the hold intact
The order is the safety property. Publishing before approval is the failure this whole page exists to prevent.
Runs when
A launch or a campaign needs an image and the request arrives as a message.
- 01
State the constraints in the thread
Sizes, channels, and what the image must not look like. A brief that only says "make a launch image" produces the last launch image.
- 02
Ask for directions, not a final
Three options cost less to reject than one finished asset, and the rejected ones are what teach the canvas your taste.
- 03
Approve in writing
Name the direction and the destination. The written approval is the record that the decision was yours.
- 04
Let the Bot deliver, then check where it landed
The Bot has the channels; the skill has the canvas. Confirm the file appeared where you meant it to.
Leaves behindAn approved asset delivered through your own channels, and a run that remembers what you rejected.
Three checks on an image agent
This is the one skill on the platform with public output, so the checks happen before delivery and after.
The hold actually held
Nothing should have been published before your pick. If an image appeared without that step, narrow the enablement list before running it again.
It survives the smallest size
Look at it as a chat thumbnail, which is where most of the audience will meet it. Launch images fail there more often than anywhere else.
The memory is real
Ask what the canvas remembers about the brief. A run that cannot recall the rejected directions will offer them again next launch.
Two bundles that make the image follow the message
A launch image should express a claim the research already supports.
What Grok Bot users ask about Felo SuperAgent
Can the Bot publish the image itself?
Through its own channels, if the approval boundary allows it. The skill prepares the asset and stops; whether anything leaves is a decision written into the Bot, not into the image.
Does the run survive the app being closed?
The canvas holds the state, so a run is not tied to an open conversation. That is what makes a long image job practical for a Bot.
Should this run on a schedule?
Usually not. Recurring text benefits from routines; recurring imagery mostly produces slightly different versions of the same asset, which looks unintentional.
What happens to the directions I did not pick?
They stay on the canvas with the reason they were rejected, which is what makes the next launch's first pass closer to your taste.
Every Felo skill for Grok Bot
Five skills written for Grok Bot. They follow the same pattern on this platform, so you can combine them in one session without re-explaining your setup.
Felo Search
Felo Slides
Felo YouTube Subtitling
Felo X SearchOne API key. A full agent toolkit.
Start with a skill, then compose the exact models, live data, knowledge, and output tools your Grok Bot workflow needs — the same key every Felo skill already uses.
Models
Use cost-efficient LLM access when your agent needs custom reasoning or generation.
- LLM API
- Chat completions
- Model access
Live information
Give agents current web and social signals instead of relying on stale context.
- Web Search
- X Search
- Web Fetch
Knowledge and output
Ground work in reusable knowledge, then turn it into shareable deliverables.
- LiveDoc
- PPT generation
- Mindmaps
Creative tools
Extend the workflow from research to visual and office-ready results.
- Image generation
- Page tools
- Office outputs