L
lead-communication-history
About lead-communication-history
A skill for answering questions about one lead’s communication history using stored message bodies and call transcripts. It resolves the lead precisely and distinguishes channels, directions, timestamps, and overlapping records.
Key features
- 01Resolve leads by name, email, or phone
- 02Retrieve communication across channels
- 03Order records chronologically
- 04Distinguish emails, messages, and calls
- 05Flag missing content and pagination
Use cases
- 01Build a timeline for one lead
- 02Verify outbound and inbound communication
- 03Find context around a specific message
Install
Copy the command and run it in a terminal.
$
npx skills add https:/ / github.com/ getnexor-ai/ skills --skill lead-communication-historyAfter installing, type/lead-communication-historyin Claude Code to confirm it loaded.
Bundled filesagents/openai.yaml
Details
From the project READMELead communication history
Goal
Answer lead-specific communication questions from the stored message bodies and call transcripts, with the correct channel, direction, and time. Do not substitute cross-lead search previews or meeting transcripts for the lead's communication timeline.
Workflow
- Resolve exactly one
lead_id.- Use the current lead scope or an explicit id when available.
- Otherwise call
list_leadswith the person's name, email, or phone. - If multiple plausible leads remain, present the distinguishing details and ask which lead; never merge their histories.
- Call
get_lead_historyfor the requested content.- For all supported communication, pass
channel: "call,whatsapp,email,sms". - For one channel, pass its exact lowercase name.
- Use
order_by: "desc"for newest-first questions andorder_by: "asc"for a chronological narrative. - Use
limitup to 200. Whenpagination.has_moreis true and the answer requires complete history, incrementoffsetand continue until it is false.
- For all supported communication, pass
- Read the content from the returned item type.
- WhatsApp and SMS:
type: "message"; readdirection,content, attachments, status, andtimestamp. - Email: read both
type: "email"rows for subject/body/delivery details andtype: "message", channel: "email"rows for conversation content. Reconcile overlapping records by direction, timestamp, provider id when present, subject, and body instead of counting the same email twice. - Phone calls:
type: "call"; readsummaryandtranscript. Acall_misseditem has no conversation transcript. - Treat
direction: "outbound"as content sent by Nexor or the operator anddirection: "inbound"as content received from the lead.
- WhatsApp and SMS:
- Use the right neighboring tool only when needed.
- Use
search_conversationsfor discovery across many leads, not as evidence for one lead: it returns bounded previews and has nolead_idfilter. - Use
list_transcriptsthenget_transcriptfor a scheduled meeting transcript. These tools do not retrieve phone-call transcripts. - Use
anchor_tswithbeforeandafteronget_lead_historywhen the user asks for context around a known message or moment.
- Use
- Answer from the retrieved evidence. Separate communication events from their content: state what happened, then what was actually said. Include channel, direction, and timestamp for material claims.
Guardrails
- Do not claim the history is complete while
pagination.has_moreis true. - Do not infer missing words from a summary, failed call, attachment, empty body, or truncated/absent transcript. State what is unavailable.
- Do not call a meeting summary a phone-call transcript, or a delivery/open event an email reply.
- Do not treat a failed or unanswered call as evidence that the phone number is invalid.
- Keep tenant and lead scope exact. Never use one lead's communication to answer about another.
- This is a read-only workflow. Do not send a reply, change a lead, or start/stop automation unless the user separately requests that action.
Output
- Lead with the direct answer or concise chronology.
- Distinguish outbound content, inbound replies, call transcript statements, and delivery-only events.
- Name gaps that affect confidence, including missing transcripts, empty bodies, attachments not represented as text, or unread pages.
Comments
Markdown supported