Rfp Clarification Q Optimizer

rfp-clarification-q-optimizer

When to Use This Skill

Use this skill whenever you need to:

• Evaluate existing clarification questions for an RFP
• Generate clarification questions from scratch based on an RFP
• Review, deduplicate, and optimize draft Q&A questions
• Create a client-facing Query Log ready to send to the client
• Identify gaps in RFP coverage that have no clarification question

Always trigger when both an RFP and a list of clarification questions are present together. Also trigger when the user says things like "generate clarification questions", "what should I ask about this RFP", "draft my Q&A", "questions for the client", "help me with clarifications", or uploads an RFP and asks anything about questions, clarifications, or gaps.

Prompts that should reliably trigger this skill:

• "Run the clarification Q optimizer on this RFP"
• "Generate a Query Log for this RFP"
• "Generate RFP clarification questions and produce the Excel and HTML report"
• "Use the rfp-clarification-q-optimizer for this RFP"
• "Create the Q&A log for this RFP - Excel format with scoring"

Overview

Produces two outputs from an RFP and optional draft clarification questions:

1. {config.branding.excel_header_title} - client-facing Excel in configured company-branded format
2. Question Assessment Report - internal interactive HTML report with PDF export, covering per-question scoring, deduplication log, gap coverage matrix, and priority classification

Shared References

This skill uses shared logic from the plugin. Read these files before proceeding:

• ../../shared/output-directory.md - config lookup, directory naming, user confirmation flow
• ../../shared/file-resolution.md - folder scanning, file role inference, confirmation protocol
• ../../shared/relationship-modes.md - relationship mode definitions, scoring adjustments, account document extraction, banner spec

Follow those files exactly for Step 0, Step 1, and relationship mode logic in Step 3.

Step 0 - Read Config and Establish Output Directory

Follow ../../shared/output-directory.md for the full protocol.

All company-specific values (branding, delivery model, frameworks, differentiators) come from config.yaml loaded through that shared flow.

Skill name for output directory: rfp-clarification-q-optimizer

Step 1 - File Resolution

Follow ../../shared/file-resolution.md for the full protocol.

File roles relevant to this skill:

Determine operating mode

• Mode A - Draft questions file is present: evaluate, deduplicate, consolidate, rewrite
• Mode B - No draft questions file: generate questions from scratch

Announce the mode to the user before proceeding.

Step 2 - Read and Analyze Inputs

Read the RFP

Extract all text, tables, and structured content. Identify:

• All stated requirements, scope items, and evaluation criteria
• Ambiguities, missing definitions, and under-specified areas
• Mandatory vs. evaluated sections
• Any published evaluation weights or scoring criteria
• Commercial, technical, operational, and SLA-related gaps

Read win themes (if provided)

Note the key themes. These inform:

• Strategic reframing of questions (dimension 18)
• Priority classification of gaps (dimension 20)

Read account documents (Existing Client modes only)

Follow ../../shared/relationship-modes.md account document extraction protocol. Build the known scope list and CFF blocklist before scoring. These feed dimension 16 question-quality penalties and dimension 18 strategic-intent framing.

Read draft questions (Mode A only)

Extract all questions. Note:

• Original numbering and grouping
• Category and sub-category labels if present
• Any context or rationale provided alongside questions

Step 3 - Analysis

Relationship Mode Scoring Adjustments

Apply clarification-question scoring adjustments from ../../shared/relationship-modes.md throughout all dimension scoring.

In practice, this means the shared rules govern how D16 and D18 behave across Prospect, Existing Client, and Existing Client New Scope modes.

Mode A - Evaluate and Improve Draft Questions

Dimension 16 - Question Quality (per question)

Flag score 1 for removal. Flag score 2 for rewrite.

Dimension 17 - Deduplication and Consolidation (document level)

• Identify duplicate or substantially overlapping questions
• Consolidate into single sharper questions
• Record: original Q numbers merged, rationale, consolidated wording

Dimension 18 - Strategic Intent / Deal Shaping (per question)

For score 3 questions where a win theme is relevant, recommend a reframed version.

Dimension 19 - Coverage of Critical Gaps (document level)

• List all major ambiguities, risks, and missing information from the RFP
• For each gap: Covered / Partially Covered / Not Covered
• For Not Covered gaps, generate a recommended question

Dimension 20 - Prioritization (per question)

Derive from RFP signals where available. Apply generic logic as fallback.

Mode B - Generate Questions from Scratch

1. Identify all RFP gaps, ambiguities, and missing specifications
2. Generate a question per gap
3. Apply relationship mode framing (deepen vs. discover, CFF blocklist, win themes)
4. Classify by priority (dimension 20)
5. Group by category and sub-category
6. Skip dimensions 16 and 17
7. Run dimension 19 as completeness check

Step 4 - Build the Final Question Set

• Remove score 1 questions
• Rewrite score 2 questions
• Apply all dimension 17 consolidations
• Apply dimension 18 strategic reframing where recommended
• Assign final category, sub-category, priority
• Renumber sequentially from 1

Default category taxonomy

Derive from RFP structure. Use defaults only where RFP provides no signals.

Step 5 - Generate Outputs

Output A - {config.branding.excel_header_title} (Excel)

Generate using openpyxl. Single sheet, no internal scoring tab.

Layout

• Rows 1-3: merged header banner, dark navy (#1F2D3D), "{config.branding.excel_header_title}" centered white bold 16pt Calibri, tagline right-aligned white 9pt italic
• Row 4: column headers, fill #2E4057, white bold 10pt Calibri
• Row 5+: alternating #DCE6F1 / #FFFFFF, 10pt Calibri #2C2C2C, thin borders #BFBFBF

Columns

Col D: wrap text. Col E: blank, fill #F2F2F2. Sort: Priority descending within each Category. Priority column NOT shown in client file.

import openpyxl
from openpyxl.styles import PatternFill, Font, Alignment, Border, Side
from openpyxl.utils import get_column_letter

wb = openpyxl.Workbook()
ws = wb.active
ws.title = "{config.branding.excel_header_title}"

ws.merge_cells("A1:E3")
ws["A1"].value = "{config.branding.excel_header_title}"
ws["A1"].font = Font(name="Calibri", bold=True, size=16, color="FFFFFF")
ws["A1"].alignment = Alignment(horizontal="center", vertical="center")
ws["A1"].fill = PatternFill("solid", fgColor="1F2D3D")

col_headers = ["Query No.", "Category", "Sub-Category", "{config.company.short_name} Query", "Response to Query"]
col_widths = [10, 20, 22, 60, 35]
for i, (h, w) in enumerate(zip(col_headers, col_widths), 1):
    cell = ws.cell(row=4, column=i, value=h)
    cell.font = Font(name="Calibri", bold=True, size=10, color="FFFFFF")
    cell.fill = PatternFill("solid", fgColor="2E4057")
    cell.alignment = Alignment(horizontal="center" if i==1 else "left",
                               vertical="center", wrap_text=True)
    ws.column_dimensions[get_column_letter(i)].width = w

thin = Side(style="thin", color="BFBFBF")
border = Border(left=thin, right=thin, top=thin, bottom=thin)
for row_idx, q in enumerate(final_questions, start=5):
    fill = PatternFill("solid", fgColor="DCE6F1" if row_idx%2==1 else "FFFFFF")
    for col_idx, val in enumerate([q["number"],q["category"],q["sub_category"],q["question"],""], 1):
        cell = ws.cell(row=row_idx, column=col_idx, value=val)
        cell.font = Font(name="Calibri", size=10, color="2C2C2C")
        cell.fill = PatternFill("solid", fgColor="F2F2F2") if col_idx==5 else fill
        cell.border = border
        cell.alignment = Alignment(horizontal="center" if col_idx==1 else "left",
                                   vertical="top", wrap_text=True)
wb.save(output_path)

Output B - Question Assessment Report (HTML)

Self-contained single HTML file. All CSS and JS inline. No external dependencies.

Color palette

Global layout

Fixed top nav bar (navy) containing:

• Left: "{config.company.short_name} | {Client} RFP - Question Assessment" in white
• Center: tab links - Overview | Questions | Gap Matrix | Dedup Log | Input Documents
• Right: "Export to PDF" button (teal background, white text)

Page background #F4F6F8. Content in white cards with subtle box-shadow. Font: Inter or system-ui, 14px base.

Tab 1 - Overview

3 stat cards (side by side):

• Overall Quality Score - large number color coded (green ≥80, amber 60-79, red <60), sub-label "Question Quality Score", small "Based on {N} questions evaluated"
• Critical Priority Questions - large count in red, sub-label "Critical Priority Questions", clickable - filters Questions tab to Critical
• Uncovered Gaps - large count in amber, sub-label "RFP Gaps Without a Question", clickable - filters Gap Matrix to Not Covered

2-panel row below cards:

Left (50%) - Critical Questions list: Q# | Category | first 80 chars of question. Each row clickable - jumps to that question expanded in Questions tab. If none: green checkmark "No critical gaps identified."

Right (50%) - Uncovered Gaps list: gap description truncated. Each row clickable - jumps to that row in Gap Matrix. If none: green checkmark "All identified gaps have coverage."

Relationship context banner (full width, below 2-panel row):

Render this as a structured status block, not a single strip. Follow the banner spec in ../../shared/relationship-modes.md for warning rules and banner states.

Tab 2 - Questions

Sticky filter bar: Dropdowns for Priority | Action | Category | D16 Score | D18 Score. Active filter badge count. "Clear all" link.

Collapsed rows (default):

[Q#] [Category] [Sub-Category] [Priority badge] [D16 ●●●○○] [D18 ●●●●○] [Action badge] [▼]

Priority badges: color coded pills. Action badges: Kept (green) | Rewritten (teal) | Removed (red) | Consolidated (gray).

Expanded row:

Original Question:
"{full text}"

Dimension 16 - Question Quality    ●●●○○  3/5  Adequate
{rationale}

Dimension 18 - Strategic Intent    ●●●●○  4/5  Strong
{rationale}

Action: Rewritten
Rewritten version:
"{rewritten text}"

Consolidated: "Merged into Q{X}" with link. Removed: reason in red text. Mode B: show D18 and Priority only, skip D16/D17.

Tab 3 - Gap Matrix

Summary bar: "{N} gaps - {X} Covered | {Y} Partially Covered | {Z} Not Covered"

Filter: All | Covered | Partially Covered | Not Covered

Table columns: # | RFP Gap / Ambiguity | Coverage | Question(s) | Recommended Question

Row left-border color: green (Covered) | amber (Partially Covered) | red (Not Covered). Coverage cell: colored pill badge. Questions cell: clickable Q# badges linking to Tab 2. Recommended Question: shown only for Not Covered rows, italic text.

Tab 4 - Dedup Log

Mode B: "Not applicable - questions were generated, not deduplicated."

Summary: "Original: {N} → After consolidation: {M} ({X} merged, {Y} removed)"

Consolidation cards:

Merged: Q3 + Q7 + Q11
Reason: {rationale}
Consolidated Question: "{text}"

Removed questions section: List of Q# with removal reason (score 1 - off-topic / already in RFP / harmful / known scope).

Tab 5 - Input Documents

Header: "Input Documents & Impact on Scoring" Sub-header: "Upload missing documents and rerun to improve question quality and scoring accuracy."

Section 1 - Documents Provided (green checkmark per row)

Table columns: Document | Type | Mandatory | What it contributed

Example rows:

• Marco_RFP_v2.pdf | RFP | Required | Source of all gaps, requirements, and priority signals
• Win_Themes.pdf | Win Themes | Recommended | Enabled D18 strategic framing across 12 questions

Section 2 - Documents Missing (color coded by mandatory status)

Table columns: Document | Type | Mandatory | Impact of Missing | Action

Mandatory status is dynamic based on relationship mode:

Row color coding in missing table:

• Required (red): light red background #FFDCE1
• Recommended (amber): light amber background #FFF2CC
• Not applicable: not shown

Impact of missing - per document:

• Draft Questions: "Running in Mode B - questions generated from scratch, not evaluated against existing draft"
• Win Themes (Prospect): "D18 strategic framing using {config.company.short_name} delivery positioning as fallback - less targeted"
• Win Themes (Existing Client): "D18 strategic scoring degraded - win theme reinforcement not possible"
• Account Plan: "Known scope list could not be compiled - D16 relationship scoring limited"
• QBR / MBR: "Recent delivery metrics unavailable - D18 partnership framing weakened"
• CFF Form: "CFF blocklist unavailable - risk of inadvertently reopening client concerns"

Action column: "Upload and rerun" (static text, not a button)

Call to action bar (teal background, bottom of tab):

To improve your score:
1. Upload the missing documents listed above to your working folder
2. Return to Claude and say: "Rerun rfp-clarification-q-optimizer with updated documents"
3. The skill will detect the new files and regenerate both outputs

PDF Export

"Export to PDF" triggers window.print() with print CSS:

@media print {
  nav, .filter-bar, .export-btn { display: none; }
  .tab-content { display: block !important; page-break-before: always; }
  .question-detail { display: block !important; }
  body { background: white; }
  .card { box-shadow: none; border: 1px solid #ddd; }
  tr { page-break-inside: avoid; }
  .tab-content::before {
    content: attr(data-title);
    display: block;
    font-size: 18px;
    font-weight: bold;
    color: #1F2D3D;
    margin-bottom: 16px;
    padding-bottom: 8px;
    border-bottom: 2px solid #1B7A8A;
  }
}

All 5 tabs print in sequence. All question rows fully expanded. Page breaks between tabs.

Step 6 - Save Outputs and Confirm

workspace/rfp-pursuits/{client}_{rfp}_{date}/rfp-clarification-q-optimizer/
  {config.company.short_name}_Query_Log_{client}_{date}.xlsx
  Question_Assessment_Report_{client}_{date}.html

Confirm to user:

Done. Two files saved:

  Client deliverable : {config.company.short_name}_Query_Log_{client}_{date}.xlsx
  Internal report    : Question_Assessment_Report_{client}_{date}.html
                       (open in browser - use Export to PDF button to save as PDF)

Final question count  : {N} across {X} categories
Questions removed     : {N}
Questions rewritten   : {N}
Questions consolidated: {N} pairs merged
Uncovered RFP gaps    : {N} (see Gap Matrix tab)

Recommendations Quality Standard

Each recommendation must be specific, actionable, and framed as suggested language.

Good: "Q4 and Q7 both ask about offshore delivery from different angles. Consolidate into: 'Are there restrictions or preferences regarding delivery from {config.company.short_name} offshore locations (India) or nearshore (Toronto, Poland, Brazil)?'"

Bad: "Some questions could be more specific."

Edge Cases

• Image-heavy RFP PDF: Note unreadable sections in Gap Matrix, flag for manual review
• No category labels on draft questions: Infer from content using default taxonomy
• Win themes not provided: D18 runs using {config.company.short_name} delivery positioning as fallback (rightshore model, AI-native SDLC, delivery maturity). Show amber warning (Prospect) or red warning (Existing Client modes) in Overview banner and Input Documents tab.
• No account docs in Existing Client mode: Apply Existing Client thresholds but note that known scope list could not be compiled - flag in Overview tab
• All questions score 4-5: Note as strong quality signal in Overview. Still run D17 and D19
• Short or high-level RFP: Flag lack of specificity as a gap in Gap Matrix