What counts as a change
Without a rule, every cosmetic edit is news. Define the threshold — a price change, a policy date, a new competitor entry — or the report will be long and unread.
Recurring reporting is the work nobody has time for and everybody needs. SuperAgent keeps the job on a canvas between runs, so each week it adds to what it already knows instead of starting from a blank document.
Felo SuperAgent on WorkBuddy
Skills → Add skill → Upload skill → felo-superagentWorkBuddy runs a bench of specialists on your desktop, in WeChat and WeCom. Skills arrive as packages with a permission list, so research becomes a capability the whole bench can call.
The reason to trust a standing report is being able to see how this week's version was built, and which step is waiting on a person.
Reads the sources agreed at setup and writes each finding to the canvas with its date.
Diffs against last week's canvas, so the report leads with what moved rather than restating everything.
A source changed the way it publishes. It asks whether to keep using it before silently switching.
Leaves a one-page draft in your workspace with the source list and the open questions attached.
Recurring work goes wrong slowly. These are the decisions that keep a standing report from becoming noise.
Without a rule, every cosmetic edit is news. Define the threshold — a price change, a policy date, a new competitor entry — or the report will be long and unread.
Pages move and feeds break. Decide now whether the run should stop, substitute, or report the failure. Silence is the one option that looks like good news and is not.
Standing jobs outlive the project that justified them. Give it a review date, so a report nobody reads gets switched off instead of accumulating.
A recurring skill is the one where the permissions conversation actually matters. Decide the cadence and the stopping condition at install time, not after the first month.
Get the felo-superagent folder from the repository. The manifest lists what a long-running job needs to keep its canvas.
Skills → Add skill → Upload skill. Check that it asks for a standing canvas and network access, and nothing about sending messages.
State the schedule, the sources, the output format, and what should happen when a source is unavailable. This is the one skill where the first prompt is really a contract.
Every Monday at 8:00, research what changed in this market,
report only the changes, and list the sources you used.The first runs tell you whether the boundary was written tightly enough.
If the second report reads like a fresh summary, the canvas is not being used and you are paying for a search every week.
Break a source on purpose. The run should say so rather than quietly produce a thinner report from old data.
Readers notice when a familiar document changes shape. If the format drifted, say so plainly in the next request.
The best first run is something you already produce on a schedule by hand.
Produce a weekly competitor brief every Monday. Report only what changed, with a link for each item.What comes backA Monday document that is short in a quiet week and visibly short, which is itself information.
Keep track of the standards consultation. Tell me when the comment period closes and what changed since the last review.What comes backA quiet watch that speaks up only when there is a date worth acting on.
A recurring report is only as good as the sources it was given.
Five skills written for WorkBuddy. They follow the same pattern on this platform, so you can combine them in one session without re-explaining your setup.
Start with a skill, then compose the exact models, live data, knowledge, and output tools your WorkBuddy workflow needs — the same key every Felo skill already uses.
Use cost-efficient LLM access when your agent needs custom reasoning or generation.
Give agents current web and social signals instead of relying on stale context.
Ground work in reusable knowledge, then turn it into shareable deliverables.
Extend the workflow from research to visual and office-ready results.