Advanced & Troubleshooting
How to Build AI Agent Skills (and When Not To)
Skills are how you teach an AI agent to work the way you work: your voice, your rules, and the tactical know-how you would otherwise repeat every single time. This walkthrough opens a real Carly account and goes through the skills actually in use, including one that drafts warm introductions end to end and one that calls an external API from inside a skill. The warm-intro skill researches both people on the web, pulls job title and company, writes one clean line of bio for each, checks the CRM for an existing connection, keeps the draft short with no corporate fluff, and appends a signature. A task that used to be dreaded now takes one message. The second half covers what skills are not. Skills are static and only get read when an agent receives a request, so anything that needs to happen on a timer is a scheduled task, and anything that needs to fire when an email arrives is a workflow. Put recurring work in tasks or workflows and put know-how in skills.
Step by step
- 01
Understand what a skill actually is
A skill is reference material an agent consults when it gets a request. It carries your voice, your decisions, and the details you'd otherwise re-explain every time you ask for the same kind of work.
- 02
Build a skill around a task you dread
The warm-intro skill is a good model: research both people, pull job title and company, write one line of bio each, check the CRM for an existing connection, keep it short, append the signature. Texting or emailing "introduce these two people" now produces a finished draft.
- 03
Put an API call inside a skill
Code is just language that executes. One skill here is an API call that generates a meme image using credentials stored in Carly integrations. Without it, the agent would have to rediscover the API, its docs, and its query structure on every request.
- 04
Save a working result as a skill
Get the agent to do the thing correctly once, then say "save that as a skill." You are capturing a run that already worked rather than writing instructions from scratch and hoping.
- 05
Use skills so your output doesn't read as AI
The rules that matter most are the negative ones: no corporate fluff, keep it short, sound like you. Those constraints live in the skill so every draft starts from your register instead of a generic one.
- 06
Know what skills are not
Skills are static and only get read when an agent receives a request. Nothing in a skill fires on its own, which is the single biggest misunderstanding people bring to them.
- 07
Use a task or a workflow when timing matters
Scheduled tasks are repeating requests on a timer. Workflows are triggered automations that fire when something happens, like an email arriving. Put recurring work there and keep know-how in skills.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."