An AI SDR agent is only as good as the definition of a good customer it is working from, which is why the fit criteria belong in a written skill rather than buried in a prompt nobody can find.
/our-icpAccount research earns its time only when it changes the first sentence you write, which means looking for what recently changed rather than assembling a profile anyone could have found.
/our-icpQualification is a claim about evidence, not a score: a deal is qualified when every criterion has a source you can quote, and unevidenced criteria are named rather than filled in optimistically.
/meddiccCompetitive positioning works when you concede what is true first: naming what a competitor does well buys credibility for everything after it, and the buyer already knows anyway.
/competitor-positioningA cold sequence works when each touch carries new information: the second email should not exist unless it says something the first one did not, and most sequences fail this test by touch three.
/email-rulesA connection note has about three hundred characters and one job: prove you know who they are, without the sentence that reveals you found them in a filtered list.
/our-writing-voicePre-meeting research has a five-minute budget, so it should return exactly three things: what changed recently, who you are talking to, and the one thing that would embarrass you not to know.
/our-icpNo. An AI SDR is an agent — it runs on a trigger and does the work. A skill is the method it applies: who counts as a fit, what disqualifies an account, and how a first email should read.
The opposite is the point. A skill is where you write down how your team actually sells, so the output sounds like your company rather than like a model with no context.
Two or three. An ICP skill, a writing-voice skill, and whatever framework you qualify with. Skills advertise themselves cheaply, so the model reaches for the right one on its own.