How claimscout/va-claim-scout works
1. What this repository is
- This is not a software application with a backend or frontend.
- It is a prompt framework distribution repo: the core “product” is written content that an external AI system loads and executes conversationally.
- The repo’s architecture is therefore centered on documentation, prompt assets, and versioned guidance, not runtime code.
2. Primary boundary: repo vs. AI platform
- The repository contains the framework artifacts.
- The actual “runtime” is an external AI chat platform such as Claude, GPT-4o, Gemini, Copilot, or Grok.
- All analysis, reasoning, and user interaction happen inside that external model; the repo itself does not process claims, store user data, or connect to VA systems.
3. Core functional artifacts
- VA_Claim_Scout_Pro.md is the main high-capability prompt framework for deeper analysis.
- VA_Claim_Scout_Lite.md is a reduced-context variant designed for lighter/free-tier AI platforms.
- These are the two operational entry points and should be treated as the main “modules” in the system diagram.
4. How the system is used
- A user either uploads or pastes one of the markdown files into an AI chat.
- The AI reads the framework and responds to the trigger phrase “Claim Scout, Help” by presenting the menu/welcome flow.
- From there, the conversation becomes an iterative analysis session where the AI helps the user reason through records, evidence, and claim strategy.
5. Data flow
- Input source is entirely user-provided: medical records, service history, symptoms, letters, exam notes, and claim questions.
- Those inputs flow into the AI chat session, where the prompt framework shapes the model’s analysis.
- Output is conversational guidance: potential conditions, secondary/presumptive claim ideas, evidence gaps, exam prep, drafted statements, and messages to representatives.
6. Main capability areas inside the framework
- The README makes clear the framework supports several analysis lanes:
- identifying potentially compensable conditions,
- surfacing secondary and presumptive claims,
- evaluating evidence strength and gaps,
- preparing for C&P exams and medical appointments,
- drafting communications to a VSO, attorney, or VA representative,
- producing buddy statements and impact statements.
- These are the main logical subsystems to reflect in an architecture diagram, even though they are implemented as prompt behavior rather than code services.
7. Important constraints and trust boundaries
- The framework explicitly does not file claims, access VA records, or replace legal counsel.
- It is positioned as an analytical aid, not a legal-service provider.
- A good architecture diagram should show a hard boundary between the AI-assisted reasoning layer and the actual VA filing/records systems.
8. Pro vs. Lite architecture split
- Pro is intended for stronger models with larger context windows and deeper reasoning.
- Lite is optimized for platforms with reduced context capacity and softer language.
- This is a compatibility split, not a feature split: both versions share the same core mission, but Lite is adapted for lower-resource AI environments.
9. Repository support files
- README.md is the main product overview and usage guide.
- CHANGELOG.md provides version history and release notes.
- LICENSE.md defines the open-source usage terms.
- VA Claim Scout – Quick Start Guide.pdf and VA_Claim_Scout.docx are packaged onboarding/reference documents.
- .github/FUNDING.yml is ancillary project-support metadata, not part of the claim-analysis workflow.
10. External ecosystem and adjacent scope
- The README references a planned companion project, SSDI Claim Scout, which is conceptually separate but based on the same prompt-framework pattern.
- It also references VA.gov for actual claim filing and notes that SSA representation is a different legal domain.
- In an architecture diagram, these should appear as external systems the user may eventually interact with, but not as components inside this repository.
11. What is notably absent
- There is no application server, database, API layer, authentication system, or build pipeline visible in the file tree.
- The repo’s “logic” lives in authored markdown content, so its architecture is primarily content distribution + external AI execution.
- That makes the repository lightweight and portable across AI platforms rather than tied to a single runtime stack.
Main entry points: README.md README · VA_Claim_Scout_Pro.md Pro prompt · VA_Claim_Scout_Lite.md Lite prompt · LICENSE.md License
Components 18
Repository assets
Versioned prompt content and supporting documentation shipped by this repo.
- README · Primary overview, usage flow, and product positioning.
README.md - Pro prompt · Main high-capability analysis prompt for deeper reasoning.
VA_Claim_Scout_Pro.md - Lite prompt · Reduced-context variant for smaller or lower-tier AI platforms.
VA_Claim_Scout_Lite.md - Quick start · Packaged reference guide for first-time use.
VA Claim Scout – Quick Start Guide.pdf - Docx guide · Packaged editable/reference version of the framework guidance.
VA_Claim_Scout.docx - Changelog · Version history and release notes for the framework.
CHANGELOG.md - License · Open-source usage terms for the repository contents.
LICENSE.md - Funding · Ancillary project-support metadata, outside the claim workflow.
.github/FUNDING.yml
AI runtime
External chat platform where the prompt framework is loaded and executed conversationally.
- AI chat · Claude, GPT-4o, Gemini, Copilot, Grok, or similar model runtime.
- Claim Scout · The interactive analysis session started after loading the prompt and trigger phrase.
- Analysis lanes · Condition identification, secondary/presumptive ideas, evidence review, exam prep, drafting, and statements.
- User evidence · Medical records, service history, symptoms, letters, exam notes, and claim questions supplied by the user.
- Guidance output · Conversational recommendations, drafts, preparation notes, and evidence gap analysis.
VA ecosystem
External systems and human actors adjacent to claim preparation and filing.
- VA records · Official VA record systems; not accessed by the repo.
- VA.gov · Actual claim filing destination and official VA process boundary.
- VSO / attorney · Representative or counsel that the framework can help the user communicate with.
- SSDI Scout · Planned companion prompt-framework project in a separate domain.
Other components
- User · Loads a prompt file and supplies claim-related material and questions.
Connections 19
- User → README: reviews · User starts with the overview and usage guidance.
- User → Pro prompt: loads · User pastes or uploads the Pro framework into an AI chat.
- User → Lite prompt: loads · User pastes or uploads the Lite framework into an AI chat.
- README → Pro prompt: points to · README presents the main prompt entry point.
- README → Lite prompt: points to · README presents the lighter compatibility variant.
- README → Quick start: supports · Quick start provides onboarding for first use.
- README → Docx guide: supports · DOCX version offers packaged reference content.
- README → Changelog: documents · Changelog tracks releases and changes.
- Pro prompt → AI chat: executes on · Prompt behavior is realized inside the external model runtime.
- Lite prompt → AI chat: executes on · Lite variant also runs in the external model runtime.
- AI chat → Claim Scout: hosts · The platform hosts the interactive Claim Scout conversation.
- Claim Scout → Analysis lanes: applies · The prompt framework drives the analytical behaviors.
- User evidence → Claim Scout: feeds · User-provided records and questions are the input material.
- Claim Scout → Guidance output: produces · Outputs are conversational recommendations and drafts.
- Guidance output → VSO / attorney: prepares · User can turn outputs into messages or talking points for representatives.
- Guidance output → VA.gov: informs · Guidance may help the user prepare, but filing happens outside the repo.
- VA records → User evidence: may inform · Official records are an external source of truth the user may compare against.
- SSDI Scout → AI chat: patterned after · Referenced as a future companion using the same prompt-framework pattern.
- License → Funding: governs · Repository support metadata is ancillary and not part of the workflow.