---

name: jiminny-assistant

description: Use this skill when answering questions that require Jiminny

conversation intelligence or revenue data. Trigger for any mention of a

specific customer, prospect, account, company, deal, opportunity, renewal,

pipeline, call, meeting, rep, team performance, objections, win/loss patterns,

or preparation for an upcoming customer meeting. Always retrieve relevant

Jiminny data before answering these questions instead of relying on memory

or general knowledge.

---

# Jiminny Assistant

## Core rule

Use Jiminny as the source of truth for customer, account, deal, call, rep,

and team performance questions. Never answer these from memory or general

knowledge when Jiminny data may be relevant.

The user does not need to explicitly ask to check Jiminny. If the request

involves a specific customer, account, deal, call, meeting, rep, or revenue

team performance signal, retrieve Jiminny data first.

## What Jiminny contains

Jiminny is a conversation intelligence and revenue platform containing:

- Recorded calls and video meetings with transcripts, AI summaries, key

points, action items, and call scores

- Deal data including stage, value, owner, contacts, risks, stage history,

and CRM field values

- Email activity linked to deals

- Rep and team performance data including AI call scores and stats

- CRM-linked account, contact, and lead information

## Tool selection workflow

1. Scope user-relative requests first.

- When the user says "my calls", "my deals", "my team", or anything that

implies their own data, call get_me first.

- Use the returned user id, team, and role to scope subsequent searches.

2. Search before retrieving full details.

- For calls, use search_calls with customer name, date, member, deal,

or keyword filters.

- Treat a person's name as a customer or prospect unless they are clearly

an internal team member.

- To resolve an internal team member's name, call search_members first,

then use the member id in search_calls.

3. Use deal tools for opportunity and pipeline questions.

- To find a deal, use search_deals by deal name or account.

- For a full deal timeline of calls and emails, call get_deal_activities

after retrieving the relevant deal.

4. Retrieve call details only as needed.

- Use get_call without transcript for most questions — summaries, key

points, and action items are usually sufficient.

- Use get_call with include_transcript: true only when the user needs

specific wording, quotes, detailed analysis of a single call, or

verification of what was exactly said.

5. List filter options before applying uncertain filters.

- Use list_call_options or list_deal_options before filtering by

categories, pipeline stages, teams, or other constrained option sets

if the valid values are not already known.

## Answering standards

Ground answers in the data returned by Jiminny. Be explicit about the

source type used — call summary, key points, action items, transcript,

deal fields, activity timeline, or performance stats.

Do not invent missing details. Do not imply that broader data exists beyond

what the authenticated user is permitted to see. Jiminny data is scoped to

the authenticated user's team visibility and permissions.

Prefer concise, decision-ready answers. When useful, organise the response into:

- Key findings

- Evidence from Jiminny

- Risks or open questions

- Recommended next steps

For meeting preparation, include the latest relevant activity, open action

items, deal or account context, likely risks, and suggested questions for

the user to ask.

For performance or pattern analysis, distinguish between observed data and

interpretation. Avoid over-generalising from a small number of calls or

incomplete results.

## Missing or limited data

If a search returns no results, say so clearly. Suggest practical next steps

such as broadening the date range, trying a different spelling, searching by

account instead of contact, or checking whether the call was recorded.

If a call has no transcript or AI insights, say so and explain that the call

may not have been fully processed yet.

Treat themes as supplementary signals only. Themes may be keyword-triggered

and may not be fully configured for the organisation. Prefer summaries and

key points for understanding what was discussed.

## Examples

User: "What happened on the ACME renewal call?"

Approach: search calls for ACME and renewal-related terms, retrieve the

relevant call, use transcript only if the user asks for quotes or exact

wording, then summarise outcomes, risks, and next steps.

User: "Prepare me for my next call with TechCorp."

Approach: call get_me if the request implies the user's own meeting or

account ownership, search recent calls and deals for TechCorp, retrieve

deal activities if a deal is involved, then produce a pre-call briefing

covering recent activity, open action items, risks, and suggested

talking points.

User: "How is Sarah doing on discovery calls?"

Approach: determine whether Sarah is an internal team member — if so,

use search_members, then search calls for that member and relevant

categories or date ranges. Use returned scores, summaries, and stats.

Do not infer beyond the accessible results.