Felo SuperAgent on Doubao WorkDoubao Work

The report needs a cover, and the cover is a job of its own

Imagery is the slowest part of a deliverable to get right and the easiest part to leave until too late. Felo SuperAgent runs that work on a persistent canvas with visible states, so the document is not blocked behind the picture on the front of it.

Built for Doubao WorkRuns on a canvasStops for your call
Get an API keySee the platform guide

Felo SuperAgent on Doubao Work

01Bundle
felo-skills/felo-superagent
02Input
A brief in words
03Comes back as
Files, plus the run that made them
04State lives
On the canvas, not in the task
The command underneathfelo superagent "cover for a quarterly industry report"
Doubao Work logo
This skill runs on

Doubao Work

ByteDance's Doubao Work · agent work platform

Read the Doubao Work guide

Doubao Work turns a request into a finished file and can run it on a schedule. A Felo skill decides what goes into that file and keeps the sources behind every conclusion.

Runs on
Personal, team, and enterprise builds
Skills arrive by
Skill panel import, or the Felo CLI
Automation
Scheduled tasks and computer use
What you get back

A run with states you can see, not a job that disappears

The useful part of a long generation is where it pauses. This one stops when taste or rights are the deciding factor, which is what keeps the output reviewable instead of merely finished.

  1. Running

    Reading the brief

    A quarterly industry report, seen on screen and printed, and it still has to read when it is 90 pixels wide in a chat preview.

  2. Running

    Three directions, first pass

    One built from the data in the report, one photographic, and one restrained type-only treatment using the report's own title.

  3. Blocked

    Waiting on your decision

    The photographic direction needs a licensing call that is not the run's to make, so it stops and says so instead of choosing for you.

  4. Done

    Exported

    Cover at presentation and print sizes, plus a 16:9 version for the title slide of the deck.

The routine

Getting a report from a draft to a finished cover

Constraints first, directions second, files last. The order is what keeps the run short and the review cheap.

Runs when

The document is nearly done and it still needs a cover, a header image, or artwork that will be seen by people outside the team.

  1. 01

    Write the brief as constraints

    Name the sizes it must survive and where it will be seen — a projector, a print copy, a chat preview. "A professional cover" gives the run nothing to check against.

  2. 02

    Ask for directions, not a final

    Three directions cost less to reject than one polished image you do not want. The first pass is for choosing.

  3. 03

    Check the smallest size first

    Look at the candidates as a thumbnail. Most covers fail there, and it is cheaper to find out before the export step.

  4. 04

    Keep the reasoning with the files

    Whoever makes the next quarter's cover needs to know which direction was rejected and why, or they will start from zero.

Leaves behindCover assets at the sizes the deliverable uses, plus a short note on the directions that were rejected and the reason.

In practice

Two prompts that keep image work short

Both hold the job at the reviewable stage instead of running straight to a finished file you may not want.

Type thisSketch three directions for the cover of this quarterly report, show them at full size and as a chat thumbnail, and stop before exporting.What comes back

Three candidates judged at both sizes, held for your pick rather than committed to a file.

Type thisThe photographic direction has a licensing problem. Replace it with something built from the report's own data and keep the same tone.What comes back

A swappable direction that avoids the rights question entirely, with the other two untouched.

Add it on Doubao Work

Import it for the canvas, or drive it through the CLI

The bundle is small and the work is not, which is why the canvas — rather than the task — is what holds the state. Import it if your build exposes a panel, or let computer use call the CLI.

Path A

Import the skill bundle

The shorter route, when your build exposes a skill or connector panel. Personal, team, and enterprise builds differ here, so check what your own account shows before planning around it.

  1. Clone the repositorygit clone https://github.com/Felo-Inc/felo-skills.git. Every bundle is a folder with a SKILL.md, and you only need the one you are installing.
  2. Import the folderBring that folder in through the skill or connector panel so report covers and product imagery is available while the task runs, not only afterwards.
  3. Confirm it is callableAsk the agent which skills it can use before you build a scheduled task around one. An imported skill that never gets called is the usual failure here.
Path B

Drive the Felo CLI

The route that works on every build. Install it once, then either run it yourself during setup or let computer use invoke it as part of a task on your machine.

  1. Install and store the keynpm install -g felo-ai felo config set FELO_API_KEY your-api-key-here
  2. Run the command by hand oncefelo superagent "cover for a quarterly industry report"
  3. Then fold it into the taskAsk the agent to call the CLI when it needs current information or a deliverable, so the same command runs inside the task instead of beside it.
Where it stops

Three things the run will not decide for you

Image generation is cheap to start and expensive to review, so it is worth knowing where the judgement stays yours.

Rights are your call

Anything photographic raises a licensing question. The run should flag it rather than resolve it, and a blocked state is the correct outcome.

It will not guess your brand

If the brief does not state the tone the team uses, the output will be plausible and generic. Bring the constraints.

Files still need filing

The canvas holds the run; the deliverable needs the exported files. Separate those two and the next quarter gets easier.

FAQ

What Doubao Work users ask about Felo SuperAgent

How is this different from the platform's own image generation?
The platform generates images as part of a deliverable. This keeps a long asset job on its own canvas with states and memory, which matters when the image goes through several rounds of feedback.
Can a scheduled task produce the images too?
It can call the CLI, but imagery usually needs a human decision. The workable pattern is a manual cover run each quarter and a scheduled text job in between.
What do I keep afterwards?
The exported files beside the document, and the note about what was rejected. The run itself lives on the canvas and is not part of the deliverable.
Does it work for video or audio assets?
The canvas is built for image work. For anything with a timeline, keep it to the platform's own generation tools and use this for the still assets around it.
More on Doubao Work

Every Felo skill for Doubao Work

Five skills written for Doubao Work. They follow the same pattern on this platform, so you can combine them in one session without re-explaining your setup.

Beyond ready-made skills

One API key. A full agent toolkit.

Start with a skill, then compose the exact models, live data, knowledge, and output tools your Doubao Work 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
Explore LLM API →

Live information

Give agents current web and social signals instead of relying on stale context.

  • Web Search
  • X Search
  • Web Fetch
Explore live data tools →

Knowledge and output

Ground work in reusable knowledge, then turn it into shareable deliverables.

  • LiveDoc
  • PPT generation
  • Mindmaps
Explore knowledge tools →

Creative tools

Extend the workflow from research to visual and office-ready results.

  • Image generation
  • Page tools
  • Office outputs
Explore platform tools →