deal-pipeline-manager

Overview

Acts as a fractional VP of Sales for small businesses: maintains pipeline health, identifies at-risk deals, prioritizes daily actions, and forecasts revenue.

Workflow

<Definition - Modes>

• Morning: "What should I work on today?" Top 5 actions only, no full analysis. Skips Steps 2-3.
• Weekly: "Weekly pipeline review." Health metrics + priorities + stale flags. Skips deep deal coaching.
• Full: "Review my pipeline" with data provided. Complete analysis + deal-by-deal coaching + forecast. </Definition - Modes>

<Definition - Data Source Cascade> Try in order, stop at first success:

1. CRM (HubSpot/Salesforce) if connected: pull open deals
2. Outlook email search: mine sent proposals, quotes, calendar meetings
3. Google Sheets: search for pipeline/deals spreadsheet
4. Excel/CSV uploaded: parse with run_python
5. Verbal input: ask user to describe deals </Definition - Data Source Cascade>

<Definition - Activity Risk Levels>

• Active: 0-7 days since last contact
• Cooling: 8-14 days
• At Risk: 15-30 days (orange flag)
• Likely Dead: 30+ days (red flag) </Definition - Activity Risk Levels>

<Definition - Skill Triggers> When a deal matches a condition, suggest the relevant skill:

• Stale deal (15+ days): suggest follow-up-cadence
• Verbal yes without contract: suggest proposal-generator
• Pipeline gap (coverage below 2.5x): suggest lead-prospector
• Won deal, no next contact: suggest outreach-sequence-builder for referral ask </Definition - Skill Triggers>

<Workflow - Morning Mode description="Quick daily priorities. Top 5 actions only." tools=[] triggers=["what should I work on today", "sales priorities", "morning priorities"]

1. [Decide] Is pipeline data available from a previous session or connected source?

   • Yes: proceed to step 2.
   • No: ask user to list their active deals briefly. Validate: Pipeline data is accessible. If fails: Present the morning mode format as a template and ask for deals.

2. [Agent] Score and rank deals by urgency: (days since activity x deal value) + stage weight. Select top 5. Validate: Exactly 5 actions produced. If fails: If fewer than 5 deals exist, include all.

3. [Agent] For each of the top 5, produce: Deal name, specific action, why-now reasoning, effort estimate (5/15/30 min). Validate: No action says just "follow up." Each is specific and actionable. If fails: Rewrite vague actions with concrete steps.

</Workflow - Morning Mode>

<Workflow - Weekly Mode description="Pipeline health metrics, stale deal flags, and priorities without deep coaching." tools=[run_python, file_write, file_read, open_in_session_tab] triggers=["weekly pipeline review", "weekly review", "pipeline health", "what deals are stuck"]

1. [Agent] Detect data source per <Definition - Data Source Cascade>. Try each in order. Validate: Pipeline data loaded from at least one source. If fails: Ask user to describe deals verbally or share a file.

2. [Agent] Parse and structure pipeline data. For each deal capture: name, amount, stage, close date, last activity, contact. Validate: At least 1 deal parsed with amount and stage. If fails: Ask user to clarify deal format.

3. [Agent] Pipeline health assessment. Calculate: coverage ratio (if target known), stage distribution, activity recency per <Definition - Activity Risk Levels>, revenue forecast (conservative/expected/optimistic). Flag stale deals (15+ days). Validate: Health metrics computed. At-risk and stale deals flagged. If fails: Provide what metrics are possible with available data.

4. [Agent] Priorities. Top 5 for today (with effort estimates), top 10 for the week. Include skill triggers per <Definition - Skill Triggers>. Skip deal-by-deal coaching scorecards. Validate: Priorities limited to max 5 daily, max 10 weekly. Skill triggers present where applicable. If fails: Trim to limits. Add missing skill triggers.

5. [Agent] Produce skill-ready outputs: proposal-ready deals table, stale deals needing follow-up table, pipeline gap brief (coverage, gap amount, ICP suggestion). Validate: All three output blocks present. If fails: Add missing outputs.

</Workflow - Weekly Mode>

<Workflow - Full Review description="Complete pipeline analysis with coaching and forecasting." tools=[run_python, file_write, file_read, open_in_session_tab] triggers=["review my pipeline", "deep pipeline review", "coach me on my deals"]

1. [Agent] Step 0: Detect data source per <Definition - Data Source Cascade>. Try each in order. Validate: Pipeline data loaded from at least one source. If fails: Ask user to describe deals verbally or share a file.

2. [Agent] Step 1: Parse and structure pipeline data. For each deal capture: name, amount, stage, close date, last activity, contact. Validate: At least 1 deal parsed with amount and stage. If fails: Ask user to clarify deal format.

3. [Agent] Step 2: Pipeline health assessment. Calculate: coverage ratio (if target known), stage distribution, activity recency per <Definition - Activity Risk Levels>, revenue forecast (conservative/expected/optimistic). Validate: Health metrics computed. At-risk deals flagged. If fails: Provide what metrics are possible with available data.

4. [Agent] Step 3: Deal-by-deal coaching. For each deal, score: decision maker identified, timeline confirmed, budget validated, competition known, next step scheduled. Produce specific next action. Validate: Every deal has a scorecard and next action. If fails: Provide abbreviated coaching for deals with limited info.

5. [Agent] Step 4: Priorities. Top 5 for today (with effort estimates), top 10 for the week. Include skill triggers per <Definition - Skill Triggers>. Validate: Priorities limited to max 5 daily, max 10 weekly. Skill triggers present where applicable. If fails: Trim to limits. Add missing skill triggers.

6. [Agent] Step 5: Produce skill-ready outputs: proposal-ready deals table, stale deals needing follow-up table, pipeline gap brief (coverage, gap amount, ICP suggestion). Validate: All three output blocks present. If fails: Add missing outputs.

</Workflow - Full Review>