sales

# Sales

Umbrella skill for sales execution: call preparation, pipeline health analysis, outreach drafting, competitive intelligence, and forecasting. Each mode loads its own references on demand. Detects the mode from the request, loads the right context, and executes the appropriate workflow.

**Scope**: Revenue-facing workflows where the user is preparing for, executing, or following up on sales activities. Use csuite for strategic business decisions. Use research-pipeline for deep multi-source research. Use voice-writer for content generation.

---

## Mode Detection

Classify the request into exactly one mode. If the request spans multiple, choose the primary and note the secondary.

| Mode | Signal Phrases | Primary Activity |
|------|---------------|-----------------|
| **CALL-PREP** | "prep me for", "meeting with", "call with", "get ready for", "before my call" | Research + agenda + questions for an upcoming meeting |
| **PIPELINE** | "pipeline review", "deal health", "stale deals", "pipeline hygiene", "which deals" | Analyze pipeline health, prioritize deals, flag risks |
| **OUTREACH** | "draft outreach", "cold email", "reach out to", "write email to", "LinkedIn message" | Research prospect then draft personalized message |
| **COMPETITIVE** | "competitive intel", "battlecard", "how do we compare", "vs competitor", "competitor research" | Analyze competitors, build positioning, talk tracks |
| **FORECAST** | "forecast", "gap to quota", "commit vs upside", "pipeline coverage", "will I hit my number" | Weighted forecast with scenarios and gap analysis |
| **CALL-SUMMARY** | "call notes", "summarize call", "follow up email", "action items from call", "what happened on the call" | Extract action items, draft follow-up, log summary |
| **RESEARCH** | "research company", "look up", "intel on", "who is", "tell me about" | Company/person research for sales context |

---

## Reference Loading Table

Load only the references required by the detected mode. Always load `references/llm-sales-failure-modes.md` for any mode.

| Mode | Load These References |
|------|----------------------|
| CALL-PREP | `references/call-prep.md`, `references/llm-sales-failure-modes.md` |
| PIPELINE | `references/pipeline-analysis.md`, `references/llm-sales-failure-modes.md` |
| OUTREACH | `references/outreach-patterns.md`, `references/llm-sales-failure-modes.md` |
| COMPETITIVE | `references/competitive-intelligence.md`, `references/llm-sales-failure-modes.md` |
| FORECAST | `references/pipeline-analysis.md`, `references/llm-sales-failure-modes.md` |
| CALL-SUMMARY | `references/call-prep.md`, `references/llm-sales-failure-modes.md` |
| RESEARCH | `references/call-prep.md`, `references/competitive-intelligence.md`, `references/llm-sales-failure-modes.md` |

---

## Workflow: CALL-PREP

**Framework**: GATHER -> RESEARCH -> SYNTHESIZE -> DELIVER

**Phase 1: GATHER** -- Collect meeting context from the user.

Ask for: company name, meeting type (discovery/demo/negotiation/check-in/QBR), attendees (names + titles), user's goal for the call, any context they want to share (paste notes, emails, prior interactions).

Accept whatever they provide. Missing fields become research targets, not blockers.

**Gate**: Company name known. Meeting type classified.

**Phase 2: RESEARCH** -- Web research to fill gaps.

Search for: company news (last 30 days), funding/leadership changes, attendee LinkedIn profiles, company product/service description, industry context. Extract: what the company does, recent trigger events, attendee backgrounds, hiring signals.

Source every company fact from search results. State 'not found' for missing data. Do not invent revenue figures, employee counts, or funding rounds that were not found in search results. See `references/llm-sales-failure-modes.md`.

**Gate**: Company profile assembled from verified sources. Gaps explicitly noted.

**Phase 3: SYNTHESIZE** -- Build the prep brief.

Produce: account snapshot table, attendee profiles with talking points, context/history summary, suggested agenda tailored to meeting type, discovery questions targeting understanding gaps, potential objections with responses.

Meeting type shapes the output:
- **Discovery**: questions > talking. Focus on qualification signals.
- **Demo**: tailored examples for their use case. Focus on technical requirements.
- **Negotiation**: objection handling, value justification. Focus on path to agreement.
- **Check-in/QBR**: value delivered, expansion opportunities. Focus on renewal signals.

**Gate**: All sections populated. No placeholder text. No fabricated details.

**Phase 4: DELIVER** -- Present the formatted brief.

Output a structured markdown brief: Account Snapshot (table), Attendee Profiles, Context & History, Suggested Agenda (numbered), Discovery Questions (5-7), Potential Objections (table: objection | response), Internal Notes.

---

## Workflow: PIPELINE

**Framework**: INGEST -> SCORE -> PRIORITIZE -> DELIVER

**Phase 1: INGEST** -- Get pipeline data.

Accept: CSV upload (preferred), pasted deal descriptions, or verbal pipeline summary. Required fields per deal: name, amount, stage, close date. Helpful: last activity date, owner, primary contact, created date.

If the user describes deals verbally, structure into a deal table before analysis.

**Gate**: Deals structured in tabular format. At minimum: name, amount, stage, close date.

**Phase 2: SCORE** -- Health assessment on four dimensions.

| Dimension | Weight | Red Flag |
|-----------|--------|----------|
| Stage Progression | 25 | Same stage 30+ days |
| Activity Recency | 25 | No activity 14+ days |
| Close Date Accuracy | 25 | Close date in the past |
| Contact Coverage | 25 | Single-threaded (one contact) |

Score each deal on each dimension. Compute pipeline health score (0-100). Identify risk flags: stale deals, stuck deals, past close dates, single-threaded deals, missing data.

Do not fabricate activity dates or deal history the user did no