| name | discovery-call-prep |
| description | Prep for a discovery call — research the account, the attendee, set hypothesis for their pains, write the question stack, define success. Use on "prep for my discovery call", "research this account before tomorrow", or "what should I ask <prospect>?". |
| allowed-tools | ["Read","Write","WebSearch","AskUserQuestion"] |
Discovery Call Prep
A great discovery call is an earned conversation, not a generic script. Prep is where the edge comes from.
Workflow
- Inputs. Account, attendee(s) + titles, date/time, source of the meeting (inbound, outbound, referral), product fit hypothesis.
- Research (web + LinkedIn + any CRM notes):
- Company: stage, funding, recent news, hiring, public pain signals
- Attendee: tenure, prior companies, what they're measured on, public content
- Peers already using us / competitors (reference ammo)
- Form 3 pain hypotheses ranked by likelihood. For each: the leading signal, the words they'd use to describe it, who else it affects.
- Write the prep doc to
./prep/<account>-<date>.md.
Prep doc template
# <Account> — Discovery — <date>
## Snapshot
- Stage/size/funding
- Recent signals (hires, launches, news)
- Why meeting now: <source + any context from prior touches>
## Attendee
- <Name> — <Title> — <tenure>
- What they likely care about: <metric/KPI>
- Prior-company pattern: <if relevant>
## Pain hypotheses (ranked)
1. <Pain> — signal: <X> — their words: "<probable phrasing>"
2. …
3. …
## Question stack
Opening: "Walk me through how you handle <workflow> today."
Going deeper:
- "What changed that put this on your plate?" (trigger)
- "Who else is affected when <current pain>?" (committee)
- "What's the cost of not solving this in <timeframe>?" (stakes)
- "What would have to be true for you to pick a solution this quarter?" (criteria)
Verification:
- "If I'm hearing you right — <paraphrase>. Is that fair?"
## Landmines (questions we don't want surfaced yet)
- Pricing before value is established
- <Competitor> comparison before pains are clear
## Desired next step
<Specific next meeting / output / access> by <date>.
## Reference ammo
- <Lookalike customer> — outcome — available for reference? Y/N
- <Case study link>
Guardrails
- Don't write > 8 questions in the stack — you won't get through them, and it becomes interrogation.
- Every hypothesis needs a signal, not just a guess. If you can't cite one, lower its rank.
- If the account is < 25 employees and you're selling to 500+, flag the mismatch and consider re-qualifying the meeting.