AskAI

AskAI is a lightweight repository for the system prompt used by the Comparify AI research assistant, a tool by The Transparency Project.

The assistant is designed to help users ask questions about public political and government-transparency data, including Congressional records, White House data, Federal Election Commission data, campaign finance records, PACs, donors, lobbying disclosures, and foreign-influence-related political activity.

Repository Contents

• full_system_prompt - The full system prompt being used by the AskAI assistant.
• compact_system_prompt.md — Compact version of the system prompt that is not current in use.
• LICENSE.md — License information for this repository.
• README.md — Project overview and contribution guidance.

Purpose

AskAI provides a structured system prompt for a nonpartisan, evidence-driven AI research assistant focused on public political records.

The assistant is intended to:

• Answer questions related to Comparify and The Transparency Project data.
• Help users interpret political and government-transparency records.
• Distinguish confirmed facts from interpretation, inference, allegation, and uncertainty.
• Refuse unrelated requests outside the scope of public political data.
• Avoid unsupported claims, defamatory allegations, and speculative conclusions.

Scope

AskAI is limited to questions involving:

• Comparify or The Transparency Project data
• U.S. politics and elections
• Congressional data
• White House data
• FEC records
• Campaign finance records
• PACs and donors
• Lobbying disclosures
• Foreign-linked political spending
• Public political and government-transparency records

The assistant should refuse questions unrelated to these areas and redirect users to an appropriate public-records research query.

Contributions

Contributions are welcome if they improve clarity, accuracy, safety, neutrality, or usefulness of the system prompt.

Good contributions may include:

• Improving prompt structure or readability.
• Strengthening source-citation requirements.
• Clarifying scope boundaries.
• Improving refusal language for out-of-scope questions.
• Adding safeguards against unsupported claims or defamatory outputs.
• Improving treatment of sensitive topics such as foreign influence, campaign finance, lobbying, and public officials.

Please avoid contributions that:

• Introduce partisan framing.
• Encourage unsupported allegations.
• Weaken sourcing requirements.
• Expand the assistant beyond its intended political-data scope.
• Add unrelated functionality or general-purpose assistant behavior.

Attribution

Assistant-ui Copyright (c) 2025 AgentbaseAI Inc.

License

See LICENSE.md for licensing details.