Skip to main content
Questy magnifying glass emblem Questy
/Ques-research /Ques-Struct /Ques-source /Ques-framework /Ques-network /Ques-report /Ques-guidelines
verified Open-Source R&D Intelligence Suite for AI Developer Agents

From Questions To Proven Answers

Bring professional scientific and technical accuracy directly into how your AI assistant thinks. Regular developer tools excel at writing code, but struggle with complex research questions. Questy embeds primary-source verification, structured taxonomies, and calibrated uncertainty directly into your agent’s reasoning loop.

download Get Skill Repository Explore All 7 Skills arrow_forward Read Evidentiary Spec
REASONING LOOP: Context → Question → Constraints → Evidence → Analysis → Conclusion → Next Action

Get Skill

Clone the Repository & Add Skills to Your AI Model

git clone https://github.com/Team-Antimatr/Questy.git

Standard Git repository clone — works across all local agent skill folders.

folder_zip Download ZIP Installation Guide open_in_new

Features

1
/Ques-research • Empirical Investigation

Systematic hypothesis testing, real-time web investigation, and empirical evidence gathering.

2
/Ques-Struct • Information Architecture

Organizing unstructured data into taxonomies, comparison matrices, schemas, and trade-off trees.

3
/Ques-source • Provenance & Citation Audit

Auditing provenance, verifying citations against primary source documents, and scoring evidentiary weight.

4
/Ques-framework • Evaluation & Decision Models

Designing reusable research methodologies, scoring rubrics, sensitivity checks, and repeatable evaluation models.

5
/Ques-network & /Ques-report • Ecosystem Mapping & Executive Delivery

Mapping stakeholder and technology dependency graphs, then synthesizing findings into traceable, decision-ready reports.

Modular Skill Architecture

Every Questy skill operates independently or hands off seamlessly across the research lifecycle. Click any command to copy its invocation prompt.

/Ques-research Core Investigation

Evidence-Led Research & Hypothesis Testing

Frames the research objective, separates stable facts from disputed claims, plans live searches, cross-checks primary sources, and evaluates competing explanations fairly.

  • • 7-step empirical operating procedure
  • • Counter-evidence & causal limitation analysis
  • • Real-time vs. underlying data date separation
/Ques-Struct Taxonomies & Schemas

Structured Matrices, Schemas & Decision Trees

Transforms chaotic notes, requirements, and multi-source findings into coherent taxonomies, comparison matrices, data schemas, and layered architectures without detaching claims from evidence.

  • • Explicit inclusion/exclusion classification rules
  • • Preserves source IDs, owners, and timestamps
  • • Exposes contradictions and missing fields
/Ques-source Provenance Audit

Source Credibility, Provenance & Citation Verification

Deconstructs claims, traces primary provenance, distinguishes original data from press-release reposts, and scores sources across 6 dimensions (Relevance, Authority, Method, Recency, Corroboration, Conflict Risk).

  • • 4-tier evidentiary hierarchy enforcement
  • • Zero fabricated DOIs, quotes, or metadata
  • • Flags dead links, paywalls, and uninspected URLs
/Ques-framework Methodology Design

Reusable Evaluation Models & Decision Rubrics

Builds operational decision frameworks with non-overlapping criteria, observable indicators, rating anchors, edge-case stress tests, and sensitivity analyses.

  • • Separates hard constraints from weighted preferences
  • • Prevents false numerical precision
  • • Includes validation plan & revision triggers
/Ques-network Graph & Ecosystem

Ecosystem, Stakeholder & Dependency Mapping

Maps how organizations, technologies, standards, and stakeholders connect. Identifies hubs, bottlenecks, single points of failure, and information flows with documented edge confidence.

  • • Separates documented edges from inferred links
  • • Generates adjacency tables & Mermaid topologies
  • • Tracks historical vs. active dependencies
/Ques-report Executive Synthesis

Decision-Ready Technical & Executive Reports

Converts verified findings into clear, audience-tailored deliverables that lead with the conclusion, connect every claim to evidence, address counter-arguments, and specify concrete next steps.

  • • Leads with executive finding & decision impact
  • • Separates evidence from interpretation
  • • Actionable recommendations with owners & risks
/Ques-guidelines Shared Constitutional Standard • 10-Point Quality Gate

Impartiality, Uncertainty Calibration & Reproducibility Guardrails

Governs all Questy skills with a 10-point final quality gate: enforcing calibrated confidence labels (Verified, Supported, Disputed, Unverified, Inferred, Unknown), temporal freshness disclosure, privacy protection, and the canonical narrative flow: Context → Question → Constraints → Evidence → Analysis → Conclusion → Next action.

View Full 7-Skill Specification →

Core Pillars & Epistemic Rigor

Questy transforms your coding assistant into a rigorous research partner by embedding institutional-grade R&D protocols directly into its reasoning loop, eliminating hallucinations and ensuring total evidentiary rigor.

Unbiased Thinking

Starts from the question—never a predetermined conclusion. Evaluates competing hypotheses and counter-evidence weighted strictly by methodological quality.

Step-by-Step System

Follows a deterministic 7-stage reasoning pipeline from scoping and verification planning to structured synthesis and proportionate action.

Real Proofs

Enforces primary-source provenance (RFCs, statutes, datasets, filings) and explicitly distinguishes verified facts from inferences or uninspected URLs.

Interactive Verification Demo

Epistemic Calibration Simulator

Select a claim status below to inspect how Questy’s /Ques-source and /Ques-guidelines calibrate confidence.

Evaluated Claim Verified

HTTP/3 uses QUIC over UDP as standardized in IETF RFC 9114 (June 2022).

Source Hierarchy Tier

Tier 1 — Primary Evidence (IETF RFC 9114 Official Specification)

Questy Evidentiary Analysis

Directly supported by inspected primary IETF standards documentation. Independent implementations (Cloudflare quiche, Chromium Cronet) corroborate protocol semantics.

Recommended Proportionate Next Action

Proceed with UDP port 443 ingress rule configuration in production load balancer.

Zero Lock-In Architecture

Layered Onto the Agent You Already Use — Not a Replacement.

Questy requires zero runtime migrations or platform lock-in. It installs as a standardized agent skill or extension across Antigravity, Claude Code, GitHub Copilot, Cursor, Codex, and Perplexity.

  • Identical declarative schema across CLI and IDE presets
  • No telemetry scraping • 100% local or private agent execution
  • Seamless fallback for offline or air-gapped repositories
code questy.config.json — Unified Skill Manifest 
{
  "$schema": "https://questy-antimatr.web.app/schema/v1.json",
  "protocol": "questy-evidential-v1.0",
  "environments": [
    "antigravity",
    "claude-code",
    "copilot-workspace",
    "cursor-rules",
    "perplexity-api"
  ],
  "verification_rules": {
    "require_primary_citations": true,
    "enforce_counter_hypotheses": true,
    "minimum_confidence_score": 0.95,
    "export_rfc_standard": "IETF-RFC-7322"
  }
}
extension Supported Agent Environments & Plugins
Antigravity Claude Code GitHub Copilot Cursor Codex Perplexity

Why Engineering Teams Choose Questy

How Questy compares against ungrounded LLM prompting and generic web-search wrappers when solving high-stakes technical and scientific questions.

Capability / Standard Raw LLM Assistant Generic Search Wrapper Questy Evidentiary Suite
Primary Source Hierarchy Mixes training memory with guesswork Treats SEO blogs equal to RFCs/papers Strict 4-tier hierarchy via /Ques-source
Counter-Hypothesis Testing Sycophantic confirmation bias Summarizes top 3 snippets without critique Mandatory competing explanation analysis
Uncertainty Calibration High-confidence tone on unverified claims Opaque inline links without confidence scope Explicit 6-tier status labels (Verified → Unknown)
Structured Output Schemas Ad-hoc prose paragraphs Bullet lists detached from provenance Traceable taxonomies & rubrics (/Ques-Struct)
Cost & Telemetry Privacy Host model tokens only Paid scraping APIs & third-party logging $0 / MIT Open-Source • 0 Extra API Keys

Frequently Asked Questions

What is Questy and how does it improve AI coding assistants? expand_more

Questy is an open-source suite of 7 modular AI skills (/Ques-research, /Ques-Struct, /Ques-source, /Ques-framework, /Ques-network, /Ques-report, and /Ques-guidelines) created by Team Antimatr. It embeds institutional-grade research protocols directly into an AI assistant's reasoning loop so the agent frames questions rigorously, checks primary sources, evaluates competing explanations, and labels uncertainty before recommending engineering decisions.

How do I install Questy in my AI model or IDE? expand_more

Run git clone https://github.com/Team-Antimatr/Questy.git and copy or symlink the ques-* skill folders into your host assistant's skills directory (for example, ~/.gemini/config/skills/ in Google Antigravity, ~/.claude/skills/ in Claude Code, or .cursor/skills/ in Cursor). See the Documentation page for step-by-step host commands.

Does Questy require paid API keys, external tokens, or telemetry services? expand_more

No. Questy is 100% free, open-source, and requires zero external API keys or paid scraping subscriptions. It runs natively inside your existing AI coding assistant using the file, search, and browser tools already permitted in your environment, with zero telemetry collection.

How does Questy prevent hallucinated citations and false certainty? expand_more

Through /Ques-source and /Ques-guidelines, Questy forbids inferring that a webpage was verified merely because a URL is known. It separates publication dates from underlying event dates, checks independent corroboration (rejecting syndicated press-release echo chambers), and requires explicit epistemic labels (Verified, Supported, Disputed, Unverified, Inferred, or Unknown).

check_circle Copied to clipboard!