Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 

Repository files navigation

RFP Battlecard

Multi-agent competitive simulation and proposal generation for any competitive opportunity.

The first AI tool that doesn't just draft proposals — it simulates the entire competitive landscape before you write a single word. Works across government contracting, academic grants, foundation funding, community development, commercial procurement, and international tenders.


The Problem

Every AI proposal tool on the market optimizes the production phase — drafting, compliance checking, formatting. But research shows that winning organizations spend the majority of their effort on capture strategy, not production. The strategy layer — competitive intelligence, win theme development, evaluator modeling — is where bids are won or lost, and no tool addresses it.

What This Does

RFP Battlecard runs a multi-agent competitive simulation before generating proposals:

  1. Profiles 8-12 competitors with specific program-level intelligence
  2. Simulates competitor bids — each competitor agent produces 3 bid variants
  3. Scores all bids through an adversarial evaluator agent using actual Section M criteria
  4. Iterates 7+ rounds — each round hardens your strategy against competitor moves
  5. Generates complete proposal drafts — all volumes, formatted to Section L specs, output as .docx
  6. Benchmarks the output — scores against 8 quality dimensions (BQB framework)

No other tool — commercial, open source, or consulting firm — does steps 1-4.

Benchmark Results

Tested against a real competitive solicitation:

Dimension Score (0-10)
Compliance Coverage 7
Evaluation Alignment 4*
Competitive Differentiation 8
Past Performance Credibility 3*
Risk Identification 9
Competitive Intelligence Depth 8
Strategic Coherence 8
Actionability 7
Total 54/80

*Low scores reflect the test company's actual qualification gaps — the tool correctly identified these as showstoppers rather than masking them. For a well-qualified bidder, projected score is 65-72.

Comparison to Alternatives

Approach Typical Score
Manual (no AI) 25-40
Generic AI (ChatGPT single-pass) 30-45
Commercial RFP tools (Responsive, Loopio) 35-50
AI-native tools (pWin.ai, AutogenAI) 40-55
Human consultant (Shipley-trained) 50-65
RFP Battlecard 54-75
Human + RFP Battlecard 70-80

Installation

This is a Claude Code plugin.

# Clone to your Claude plugins directory
git clone https://github.com/YOUR_USERNAME/rfp-battlecard.git ~/.claude/plugins/rfp-battlecard

Then restart Claude Code. The skill auto-registers.

Usage

1. Drop Everything in One Folder

Dump all your files into a single folder — RFP docs, org profile, resumes, past performance, certifications, whatever you have. Flat or organized into subfolders — doesn't matter.

my-rfp-folder/
├── RFP-2026-0042.pdf           # The actual RFP (required)
├── Amendment-001.pdf            # Amendments, Q&A
├── company-overview.docx        # Your org profile
├── john-smith-resume.pdf        # Key personnel
├── case-study-HUD-2024.pdf      # Past performance
├── sam-registration.pdf         # Certifications
└── teaming-agreement.docx       # Partners/subs

2. Run a Command

Full Bid Pipeline:

/rfp:bid /path/to/my-rfp-folder

Auto-classifies all files → competitive intelligence → multi-agent simulation (7 rounds) → full proposal drafts → .docx export → quality benchmark. After your first bid, the skill offers to save your org profile so future bids only need the RFP document.

Time: 2-4 hours | Output: 25+ documents in organized directory structure

Quick Evaluation (Go/No-Go):

/rfp:evaluate /path/to/my-rfp-folder

Fast assessment: auto-classify files, identify gate requirements, profile top 5 competitors, estimate win probability, recommend Go/No-Go.

Time: 15-30 minutes | Output: Single evaluation summary

Benchmark Existing Bid:

/rfp:benchmark /path/to/bid/folder

Score any bid package (yours or a competitor's) against the BQB framework.

Time: 15-20 minutes | Output: Validation report with scores and recommendations

3. Reuse Your Org Profile

After your first bid, the skill saves your org profile to ~/.rfp-battlecard/org-profile/. On future bids, just drop the RFP PDF in a folder and run — your org data loads automatically.

Output Structure

bid/
├── 00-competitive-intelligence/    # Competitor profiles, market brief, SWOT
├── 01-simulation/                  # 7+ rounds of competitive simulation
├── 02-proposal/                    # Complete proposal volumes (I-IV)
├── 03-supporting/                  # Win themes, risk register, tech concepts
├── 04-validation/                  # BQB benchmark report
└── docx/                           # All documents as .docx (mirrored structure)

How It Works

Agent Architecture

Agent Role
Evaluator Government SSEB evaluator — scores against Section M with line-by-line compliance checks
Competitors (3-5) Top competitors, each producing 3 bid variants based on real capabilities
Client Your company — generates and refines 3 strategies per round

Simulation Rounds

Round Focus
1 Baseline bids — all agents produce initial strategies
2 Gate compliance — creative past performance framing
3 Discriminator sharpening — what competitors can't match
4 Task order positioning — which work you can realistically win
5 Small business strategy — socioeconomic optimization
6 Risk mitigation — address every evaluator-identified weakness
7 Final optimization — consolidated strategy with Go/No-Go
8-10 Stress testing — adversarial rounds targeting your weaknesses

What Makes This Different

Feature Commercial Tools RFP Battlecard
Competitor bid modeling No Yes — 3-5 competitors × 3 variants
Adversarial evaluator scoring No Yes — Section M line-by-line
Iterative strategy refinement No (single-pass) Yes — 7+ rounds
Ghost theme generation No Yes — per competitor
Go/No-Go decision framework Basic checklist Quantitative multi-gate scoring
Creative past performance framing Template-based Strategy-driven narrative
Dual-track recommendations No Yes — prime + sub backup
Win probability estimate No Yes — with confidence interval

Adapts to Any Sector

Sector What It Does
US Federal Government Full FAR compliance, Section L/M parsing, OCI analysis, SB plan, clearance assessment
State/Local Government Simplified compliance, DBE/MBE/WBE goals, local preference analysis
Defense/Intelligence Security gates, ITAR, classified requirements, DD254
Academic Grants (NSF/NIH/DOE) Intellectual merit, broader impacts, PI qualifications, budget justification
Foundation Grants Theory of change, impact metrics, funder alignment, sustainability plan
Community Development (CDBG/HUD) Needs assessment, equity analysis, stakeholder engagement, match funding
Commercial/Corporate ROI analysis, SLA commitments, pricing strategy, vendor references
International (USAID/World Bank/EU) Country context, local partners, donor priorities, MEL frameworks
SBIR/STTR Innovation narrative, commercialization plan, technical objectives
Quick Eval (any sector) Go/No-Go recommendation in 15 minutes

Requirements

  • Claude Code CLI
  • Python 3 with python-docx (pip3 install python-docx)
  • poppler for PDF extraction (brew install poppler)

Market Context

  • 68% of proposal teams now use generative AI (doubled from 34% in 2023)
  • Average RFP win rate: 45% (2025) — top firms reach 70%+
  • Price-to-Win studies cost $20K-$300K and take weeks
  • This tool compresses that to hours
  • Currently a Claude Code plugin — standalone (platform-agnostic) version planned

License

MIT

Contributing

PRs welcome. Key areas for improvement:

  • FPDS/USASpending.gov data integration for automated competitor research
  • Price-to-Win modeling with historical rate analysis
  • Agency-specific evaluation pattern libraries
  • Integration with SAM.gov opportunity tracking

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors