Make a launch image for Thursday's announcement. Show me three directions in the thread before anything is sent.What comes backThree candidates posted back into the conversation, with nothing published until you pick one.
Felo SuperAgent on OpenClaw 2.0
OpenClaw 2.0
OpenClaw 2.0Ask for the image in the thread, approve it before it goes anywhere
Your assistant already handles the conversation and the delivery; the missing piece is the asset. Felo SuperAgent runs that work on a persistent canvas, so the run survives the chat session and stops where your approval boundary begins.
- Installclawhub install felo-superagent
- InputA brief in the thread
- Comes back asFiles, plus the run that made them
- DeliveryYour approval rules decide
Install from the registry
clawhub install felo-superagentThis skill runs on
Read the OpenClaw 2.0 guideOpenClaw 2.0
Open-source personal assistant · release v2026.8.1
OpenClaw treats a skill as something with a lifecycle — install, review, then enable per agent — and answers wherever you already chat, so the reply has to hold up on its own.
- Registry
- ClawHub
- Review
- Skill Workshop keeps proposals and checks
- Reach
- WhatsApp, Telegram, Slack, Discord, native apps
A run that keeps its state while the chat moves on
Chat sessions are short and images take longer than one, which is the whole reason to keep this on a canvas. The run pauses for the decisions that are not its to make.
- Running
Reading the brief from the thread
A launch announcement image, posted to three channels, and it has to survive a chat thumbnail before it survives anything else.
- Running
Three directions
One built from the product screenshot, one typographic, and one using the launch date as the whole composition.
- Blocked
Held at the approval boundary
Two directions are ready to post. Nothing leaves until you approve, which is the rule you set for those channels and not a limitation.
- Done
Delivered and archived
The approved image in the channels you chose, with the run kept on the canvas so the next launch starts from the same rules.
From a sentence in the chat to an approved asset
Constraints, directions, approval, delivery. The approval step is not a formality here; it is the part that makes the rest safe.
Runs when
A launch, a release, or a campaign needs an image and the request arrives as a message rather than a brief.
- 01
Turn the sentence into constraints
Ask what sizes it will be seen at and where it will be posted. A chat thumbnail and a presentation slide impose different limits, and the run needs both.
- 02
Request directions, not a final
Three options cost less to reject than one finished image. This is also the point where the canvas remembers what you turned down.
- 03
Approve explicitly, in the thread
The approval boundary is the feature. Say which one goes and where, so the record shows a decision rather than an assumption.
- 04
Let the channels do the delivery
OpenClaw handles posting under your rules, so the skill only has to produce an asset worth sending.
Leaves behindAn approved asset delivered through your own channels, plus a run on the canvas that makes the next launch faster.
Add it on OpenClaw 2.0Install, review, then decide which agents may generate images
Image generation is the skill here with the most visible output, so the enablement list is the step worth taking slowly. Read the review notes before you widen it.
From the official registry
clawhub install felo-superagentSkill Workshop, before anything is live
- What the skill proposesThe brief, the generation call, and where the files are written. Read this one properly if any agent of yours can post to a public channel.
- What ran against itThe checks OpenClaw recorded for the proposal, kept with the skill so the decision to enable it stays auditable.
- What stays on recordApplied skills remain in the Workshop with their check history, which is the difference between an install you can audit and a folder somebody copied in.
Who gets it
Main chat agentEnabledWhere the brief gets written, usually mid-conversation, which is the right moment to capture it.
Channel publishing agentReview firstEnable it here only once your approval rules for that channel are the ones you want, because this is the agent that can send an image out.
Scheduled briefingsNot neededRecurring text work rarely needs new imagery each run, and reusing the approved assets usually looks more deliberate anyway.
Two bundles that make the image a decision rather than a guess
An image for a launch should follow from the message, not precede it.
Two prompts that fit the way people actually ask
In a chat, the brief arrives as a sentence. Both of these keep the run at a size worth reviewing.
The typographic one is closest. Make the date bigger and hold it for my approval again.What comes backA revised candidate sitting at the approval boundary, with the previous version kept for comparison.
Three checks before the asset is published
Channel posts are public and permanent, so these happen before delivery rather than after.
The approval boundary held
Nothing should have been sent before you chose. If an image appeared in a channel without that step, the enablement list needs narrowing before anything else.
It reads at thumbnail size
Look at it in the chat where it will appear. Most launch images fail there, and it is the size most people will actually see.
The run kept its memory
Ask what the canvas remembers about the brief. A run that cannot recall the rejected directions will offer them again next time.
What OpenClaw users ask about Felo SuperAgent
Can the agent post the image by itself?
Only within the approval rules you set for that channel. The skill produces the asset; OpenClaw decides what may leave, which is the arrangement that makes an image agent safe to run.
Where do the files live?
The run lives on the canvas and the exported files go where your setup puts them. Keep the approved assets together so the next launch can reuse them instead of regenerating them.
What happens in a long chat with many revisions?
The canvas holds the state, not the thread. That is the reason to use it here: a session that ends does not take the run with it.
Is it worth enabling on a scheduled agent?
Rarely. Recurring text benefits from a schedule; recurring imagery usually does not, and reusing approved assets tends to look more consistent than generating new ones each week.
Every Felo skill for OpenClaw 2.0
Five skills written for OpenClaw 2.0. 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 OpenClaw 2.0 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