The Bot's thread
ConversationThe default and often the best. You can ask a follow-up question of the same method without re-explaining anything.
Felo Search on Grok Bot
Grok Bot
Grok BotThe brief your Bot writes is only as good as the dates on it
A Bot that owns a recurring brief will happily produce it every week, which is exactly why the sourcing has to be built into the method. Felo Search puts a page and a date behind every claim, so the routine reports changes instead of restating last week.
- Core units
- Bot · skill · routine
- Skill comes from
- A run you already did
- Comes back as
- Claim, source, date
- Credential
- FELO_API_KEY
The request that saves the method
Save this process as a skill called "Weekly market brief". Source and
date every item, report only changes, and hold anything unverified
in a separate section.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
A brief where every line carries the page it came from
On a scheduled routine the output has to stand on its own, because nobody watched it being made. The dates are what let a reader tell this week's finding from a sentence the method has been repeating since March.
FindingSourcePublished
The regulator opened a consultation on feesregulator consultation page2 days ago
The competitor's entry tier gained a free seattheir pricing page5 days ago
Q3 filings are due on the 15thfiling service notice1 week ago
The grant cap is unchanged this cyclefunding body page3 weeks ago
A rule change is expected in 2027draft only, not in forceto confirm
Bake the change rule into the saved skill: report what is new since the last run, keep unverified items in their own section, and say so plainly when a week produced nothing.
Add it on Grok BotThis skill is something your Bot learns, not something you copy
Grok Bot builds skills out of work that has already been done, which means the first run matters more than the install. Do the job once, properly, then ask for it to be recorded.
- Run
Give one Bot a real job
Create a Bot with a narrow ownership — "Weekly Market Brief", not "General Helper" — and describe the approval boundary in the description. Report only, never contact anyone.
- Save
Ask it to record the method
Ask the Bot to save the process as a skill. A good saved method names its inputs, its validation step, the output format, and what it should do when a source is unavailable.
- Enable
Turn the skill on where it is needed
Skills are available across your Bots, but a Bot still needs the relevant plugin or login to use one. Enable it per Bot, then confirm the answer carries sources.
- Schedule
Attach a routine, then read the first run
Attach a routine with a schedule and a time zone, then read the first run as if you were the audience. Confirm that a quiet week produces a short brief rather than a padded one.
The brief that gets shorter as the method improves
A recurring brief earns its place by being read. These four moves are what keep it from growing every week.
Runs when
A market, a policy area, or a competitor set that somebody on the team has to keep an eye on.
- 01
Run it once by hand and read the sources
Before it becomes a routine. A scheduled job that produces unchecked claims repeats the mistake weekly and calls it consistency.
- 02
Save the method with the change rule in it
The instruction that matters is "report what is new since the last run". Without it, week eleven is as long as week one and nobody reads either.
- 03
Keep unverified material in its own section
Drafts, rumours and unsourced claims belong together, visibly. When one firms up, that movement is the news.
- 04
Read the first three runs yourself
The routine is only trustworthy once you have seen it handle a quiet week and a week where a source was down.
Leaves behindA short brief of what changed, with dates and sources, a separate section for what could not be confirmed, and an honest note when nothing moved.
Three places the brief can land, all inside the Bot's own loop
The Bot has a conversation and a schedule; deciding the destination before you attach the routine saves a rebuild later.
Routine output
Scheduled runRuns while the app is closed. Confirm the owning Bot, the time zone, and what happens when a source is unreachable.
The team's channel
Shared summaryIf the brief is for other people, check the approval boundary in the Bot description before it starts sending anything anywhere.
Two requests that produce a method worth saving
The first run is the template for the skill, so it is worth asking for exactly what you want to keep.
Research the three competitors on this list: pricing, packaging, and launch news from the last seven days. Source and date every item, and finish with a one-paragraph summary.What comes backA brief with a consistent shape, which is what the saved skill will reproduce every week afterwards.
Save this process as a skill called "Weekly market brief". Include the source format, the change rule, and the rule that nothing gets posted without my approval.What comes backA reusable method attached to the Bot, with your boundary written into it rather than remembered by you.
Three things a saved method will not fix
Automation preserves whatever the original run did, including its bad habits.
A saved guess is still a guess
If the first run produced claims you could not check, the skill will produce them every week with increasing confidence. Fix the method before you schedule it.
A routine needs a failure rule
Sources go down. Unless the skill says what to do, a bad week produces a brief built on old data with today's date on it.
Enablement is per Bot
A skill saved in one Bot is not automatically in use everywhere. Turn it on where it belongs, and leave it off where an unsourced answer would do damage.
What Grok Bot users ask about Felo Search
Do I need to install something first?
The saved method calls the Felo CLI, so install it once and store the key. After that, the skill is just the procedure your Bot follows.
Can I use a packaged skill instead of saving my own?
Yes — Settings → Plugins → Yours takes the packaged route, and it is faster. Saving your own is better when the output format matters to you, because the format is part of the method.
What stops the Bot from posting something on its own?
The approval boundary you write into the description and the skill. Naming it explicitly is the difference between a research Bot and a Bot that speaks for you.
How do I know the routine is still working?
Read a run a month. Methods rot quietly: a source changes shape, and a brief that looks fine starts citing a page that no longer exists.
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 Slides
Felo YouTube Subtitling
Felo X Search
Felo SuperAgentOne 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