The thread
OutlineReview it in conversation, where changing your mind is a message rather than a regeneration.
Grok Bot has its own working computer, so producing the file is not the problem. Felo Slides gives you the argument first, as an outline built from the research the Bot already gathered, so the deck you approve is the one you asked for.
Save the process as a skill called "Monthly briefing deck". Outline
first, one line per slide, and wait for my approval before exporting.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.
The useful method here is not "make slides" — it is your house shape: outline first, approval, then export. Run it once with that shape and save it.
Give one Bot the briefing job and run the research inside it, so the deck has something real to summarise rather than a topic.
Ask for the outline, argue with it, approve it, then export. Save the process afterwards and say explicitly that the outline comes before the export.
Enable the skill for the Bots that present work, and keep it off the ones that only answer questions. A deck is a deliverable, not a reply.
Attach a monthly routine for recurring briefings. Keep the headings stable so the audience reads the figures that changed rather than the redesigned slides.
The point of asking for the outline first is that it is cheap to change. Once the Bot has exported a deck, every correction costs a regeneration, and the reviewers have already seen a version you did not intend.
Keep the source list in the speaker notes when the deck is exported. The audience reads the number; the presenter is the one who gets asked where it came from.
Four steps and one gate. The gate is the part that makes a Bot-generated deck safe to present.
A recurring briefing that somebody has to present, or a one-off deck built from research the Bot has already done.
A deck built from a topic is a template with words in it. Build it from the findings instead, with the sources still attached.
Six headings, one line each. This is the last moment when changing the argument costs nothing and embarrasses nobody.
Say which slides go and which do not. An approval that consists of silence is how a deck ends up presenting a recommendation you never made.
The Bot has the computer and the file handling; the notes are what let the presenter survive the first hard question.
Leaves behindA deck that follows the argument you approved, with the sources carried into the speaker notes.
The Bot can save and share files; what it should not do is guess where a deck belongs.
Review it in conversation, where changing your mind is a message rather than a regeneration.
The Bot's own computer writes the file. Naming the folder up front saves hunting for it later.
Export last, after the argument settles, and keep the approval boundary on anything that leaves the account.
A generated deck is confident by default, so the checks are about where each claim came from.
Compare the two. Drift between them is the most common failure, and it is invisible unless you look.
A number without one is a number from an unknown quarter. The research behind the deck should still show its dates.
Read the last two slides and check that they say what you decided, not what the method defaulted to.
The deck sits at the end of a chain, and the earlier links decide how much of it holds up.
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.
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.
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.