v14-grounded-research

Produces evidence-grounded research syntheses with retrieval discipline, source ranking, and citation control.

Updated May 22, 2026
One-click install
npx skills add https://github.com/handaeho/Harness_Engineering --skill v14-grounded-research-handaeho
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: v14-grounded-research
Source: https://github.com/handaeho/Harness_Engineering/tree/main/prompt-stack/_legacy/_cold_storage/legacy_older_versions/v14/codex/skills/v14-grounded-research
Command: npx skills add https://github.com/handaeho/Harness_Engineering --skill v14-grounded-research-handaeho

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Research and analysis tasks often produce answers that overstate what the evidence supports, hide source conflicts, or rely on stale information. This Skill enforces a disciplined workflow for document-grounded investigation so that every claim is backed by retrieved evidence, freshness boundaries are explicit, and unresolved conflicts are surfaced instead of smoothed over. ## Core Features & Use Cases - Evidence-Target Discipline: Defines exactly what claim needs support, what freshness level matters, and what claim strength is justified before any retrieval begins. - Bounded Search and Retrieval Path Selection: Chooses the cheapest sufficient retrieval route (direct, hybrid lexical + semantic, or iterative) and stops when decision leverage collapses. - Source Ranking and Conflict Handling: Ranks sources by authority, freshness, scope fit, and directness, and preserves or weakens claims when sources materially disagree. - Grounded Synthesis with Citations: Separates source-supported claims from inference, cites load-bearing claims, and states what remains unsupported. - Use Case: Ask for a comparison of two API providers' current pricing and rate limits. The Skill retrieves current sources, flags conflicting documentation, and delivers a synthesis with citations and an explicit statement of what could not be verified. ## Quick Start Use the grounded research skill to investigate this question with cited sources and an explicit statement of any conflicting or unsupported claims.

Frequently Asked Questions about v14-grounded-research

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get evidence-backed answers with citations from an AI?

Define the exact claim that needs support before retrieving, gather only the minimum supporting passages, and cite load-bearing claims while preserving source-to-claim mapping. This Skill enforces that workflow and states what remains unsupported.

How to handle conflicting sources in research synthesis?

Rank conflicting sources by authority, freshness, and scope fit, then resolve cleanly when possible. When the conflict is outcome-changing and not safely resolvable, preserve it explicitly and weaken the claim strength rather than averaging incompatible facts.

When should I use grounded retrieval versus answering from context?

Use grounded retrieval when the answer depends on documents, fresh information, or verifiable citations, or when source conflict could change the answer. Skip it for self-contained local tasks or trivial questions already supported by active context.

Does this approach check information freshness?

Yes, when the task is time-sensitive it makes the freshness boundary explicit, prefers the most current source matching scope, and never implies currentness that was not actually checked. Non-time-sensitive tasks skip unnecessary freshness checks.

What are the limitations of deep research mode?

Deep research mode uses bounded rounds of sub-questions, retrieval, and gap checks, stopping once evidence coverage is sufficient. It avoids giant research plans for small questions and repeated low-yield searches after the answer is strong enough.