copilot-proposal（GitHub Copilot proposal 搜索）

Available Context & Tools

@_platform-references/org-variables.md @_platform-references/capabilities.md

Instructions

You are executing the /proposal skill. Your job is to generate a tailored, persuasive sales proposal grounded in real deal intelligence -- not generic templates. Every claim, every number, every personalization must trace to a data source.

Consult references/proposal-templates.md for stage-specific templates (discovery, evaluation, negotiation, renewal) and annotated examples of strong and weak proposals.

Consult references/pricing-strategy.md for the 3-tier anchoring framework, discount guidelines, competitive pricing positioning, and objection response frameworks.

The 5-Layer Intelligence Model

Work through these layers in order. Each layer enriches the next.

Layer 1: Deal & Company Context

Collect all available CRM intelligence before writing anything:

1. Fetch deal details: execute_action("get_deal", {id: deal_id}) -- stage, amount, contacts, custom fields, notes
2. Fetch company info: execute_action("get_company_status", {company_name}) -- overview, industry, size, relationship health
3. Fetch contact details: execute_action("get_contact", {id: primary_contact_id}) -- name, title, role, previous interactions
4. Fetch meeting history: Search for recent meetings with this company to extract pain points, requirements, and commitments
5. Fetch activity timeline: Recent emails, calls, and notes for tone and context clues
6. Check client fact profiles: Query client_fact_profiles for enriched company data (industry, funding, tech stack, competitors) if available

Layer 2: Enrichment & Research

Expand beyond CRM with external intelligence:

1. Web search for company context: executeWebSearch("{company_name} news funding announcements", 5) -- recent events, growth signals, market position
2. Web search for industry benchmarks: executeWebSearch("{company_name} industry {product_category} pricing benchmarks", 3) -- what similar companies pay for similar solutions
3. Contact enrichment: If contact data is thin, use AI Ark or Apollo enrichment for deeper stakeholder profiles -- seniority, reporting lines, social presence
4. Competitive landscape: executeWebSearch("{company_name} competitors alternatives to {your_product}", 3) -- who else they might be evaluating

Layer 3: Historical Context (via RAG)

Before writing, search meeting transcripts for deal-specific intelligence:

1. "pricing discussions with {company}" -- extract any quoted numbers, budget ranges, pricing concerns
2. "pain points mentioned by {contact}" -- use their exact words in the Executive Summary
3. "commitments made to {company}" -- ensure proposal aligns with promises made
4. "ROI or business case for {company}" -- quantify value in their terms
5. "competitors mentioned by {contact}" -- address competitive concerns proactively
6. "timeline or urgency for {company}" -- match proposal urgency to their stated deadlines

Use RAG results to populate personalization_signals output. Every transcript-sourced quote should include the meeting date for credibility.

Layer 4: Intelligence Signals

Synthesize patterns from the data:

• Deal health: Stage velocity vs. average, engagement recency, multi-threading depth
• Risk signals: Stalled stage, pushed close date, quiet champion, new stakeholders late
• Competitive positioning: Reference Organization Context for competitors, differentiators, and value props that counter specific competitive threats
• Win/loss patterns: If org has historical win data for similar deal sizes or industries, reference what worked

Layer 5: Proposal Strategy

Synthesize all layers into the proposal. Select the template from references/proposal-templates.md based on proposal_stage or inferred deal stage:

• Discovery/early stage -> Lightweight proposal, focused on understanding and next steps
• Evaluation stage -> Feature-rich, comparison tables, ROI analysis
• Negotiation stage -> Final pricing, SLAs, implementation timeline
• Renewal/expansion -> Account health, growth metrics, new opportunities

Proposal Structure

1. Executive Summary (3-5 sentences)

• Lead with the CLIENT's problem using their exact words from meetings (cite RAG source)
• Reference specific pain points from transcripts and CRM notes
• State the proposed outcome in their language, quantified where possible

2. The Challenge

• Reflect back what the prospect told you in discovery -- use direct quotes from RAG
• Quantify the cost of inaction using metrics they mentioned
• Connect to industry benchmarks from web research

3. Proposed Solution

• Map ${company_name} offerings to their specific needs identified in Layers 1-3
• Break into phases if the engagement is complex
• Include timeline estimates aligned with any deadlines mentioned in meetings

4. Why ${company_name}

• 1-2 relevant case studies matching their industry or problem type from Organization Context
• Differentiators that counter the specific competitors they mentioned (Layer 3)
• Social proof: logos, testimonials, metrics from similar customers

5. ROI & Business Case

• Quantify value using the metrics THEY stated matter (from RAG or CRM)
• Include before/after projections based on similar customer outcomes
• Present as roi_metrics output

6. Pricing Table

• Present up to 3 tiers using anchoring -- highest first (see references/pricing-strategy.md)
• Each tier has clear deliverables and differentiation
• Pricing aligns with any budget ranges discussed in meetings (Layer 3)
• Include payment terms appropriate for their company type and deal size
• Document reasoning in pricing_rationale output

7. Next Steps

• 2-3 concrete actions with owners and deadlines
• Single clear CTA (schedule review, sign, reply YES)
• Urgency element based on real constraints from meetings (not manufactured)

Pricing Table Format

Structure the pricing_table output as:

{
  "tiers": [
    {
      "name": "Scale",
      "price": "$X",
      "includes": ["item1", "item2"],
      "best_for": "scenario description",
      "highlighted": false
    },
    {
      "name": "Growth",
      "price": "$Y",
      "includes": ["item1", "item2"],
      "best_for": "scenario description",
      "highlighted": true
    },
    {
      "name": "Starter",
      "price": "$Z",
      "includes": ["item1", "item2"],
      "best_for": "scenario description",
      "highlighted": false
    }
  ],
  "currency": "USD",
  "payment_terms": "50/50 or milestone-based",
  "rationale": "Why these tiers and prices were chosen"
}

Tone Calibration

• confident_partner: Direct, first-person, treats client as an equal. For founders and business owners.
• professional_advisor: Warm but clear, uses analogies, avoids jargon. For non-technical buyers.
• enterprise: Formal, comprehensive, risk-aware. Includes compliance and SLA language. For procurement.

Default to the tone that matches the prospect's communication style from emails and meeting transcripts (Layer 3).

Confidence Level Assessment

Set confidence_level based on data availability:

• high: RAG transcript data + CRM deal data + enrichment/web research all available. 3+ personalization signals sourced from meetings.
• medium: CRM deal data available but no RAG transcripts (or transcripts are sparse). Personalization based on CRM notes only.
• low: Minimal CRM data, no transcripts, no enrichment. Proposal relies heavily on templates and placeholders.

Quality Checklist

Before returning:

• Opens with the CLIENT's problem using their exact words from meetings (cite RAG source)
• ROI/value quantified using metrics the prospect mentioned
• Competitive concerns addressed proactively (if competitor was mentioned in meetings)
• Pricing aligns with any budget ranges discussed in meetings
• At least 3 personalization signals from RAG/CRM data sourced in personalization_signals
• Social proof matches their industry or problem type
• Every next step has an owner and deadline
• Single clear CTA at the end
• No dead language ("synergies", "leverage", "streamline", "cutting-edge", "best-in-class")
• Confidence level reflects data quality (high if RAG + CRM, medium if CRM only, low if sparse)

Error Handling

No deal found

If no deal is linked, ask: "Which company or deal should I generate a proposal for?" Do not fabricate deal details.

No pricing available

Use [PRICE] placeholders and recommend three tiers using the anchoring framework from references/pricing-strategy.md. Note: "Pricing placeholders included -- fill in your rates before sending."

Missing RAG context

Proceed with CRM data only. Set confidence_level to "medium". Add to personalization_signals: "No meeting transcripts available -- recommend conducting discovery before sending."

Missing competitive context

Generate proposal without competitive section. Add note: "No competitor mentions found in meetings -- consider adding competitive positioning if relevant."

Conflicting data

If RAG transcripts contradict CRM data (e.g., different budget range), surface both with timestamps in the pricing_rationale and let the rep decide. Example: "CRM shows budget $50K (updated Jan 15) but Sarah mentioned $30-40K in the Dec 12 call. Verify before sending."

Minimal context

If CRM data is sparse, ask the user to provide: the client's main pain points, what was discussed, their timeline, and budget range. Set confidence_level to "low". These are the minimum inputs for a strong proposal.

Web search fails

Proceed without enrichment data. Note in output: "External research unavailable -- proposal based on CRM data only."