3D Deep Research: Evidence-Driven Reports for Agents

Where this fits

A developer asks their agent for “competitive analysis of Acme Cloud” and gets a five-bullet summary with no sources. When research questions need traceable conclusions, not vibes, this skill steps in: it turns the request into a structured investigation with an evidence ledger, timelined causality, force-field analysis, and a formal report in Markdown, HTML, and PDF.

Why agents benefit

• Builds two ledgers: a source ledger and a claim ledger, so every statement can be traced back to a reference or explicitly marked as inference.
• Applies X/Y/Z analysis: X timeline causality, Y key-snapshot force fields, Z internal mechanisms, and forces a mechanism-level judgment instead of a summary.
• Uses question-driven chart configuration: charts appear only where they answer a question, not as decoration.
• Writes with layered density: body text stays readable; rigor moves to appendices without deleting evidence.
• Delivers standard output templates, so reports are consistent across runs and easy to share.

Practical scenarios

Competitive due diligence for a startup

Before an investor meeting, the agent researches a competitor: timeline of pivots, forces in the current market, and the internal operating model that explains why they are winning. The output is a 20-page PDF with clickable sources, not a list of features.

Tracing a technology standard's history

For a platform team evaluating Rust vs Go, the agent builds a causal timeline of language evolution, zooms into the 2016 and 2022 pressure points, and maps each ecosystem's governance mechanism. Decision makers can verify each claim.

Industry market review with evidence trail

A policy researcher asks for “how battery recycling became a contested market.” The skill separates commodity flows, regulation changes, and company strategies into X/Y/Z axes, then produces a report with charts and full evidence appendices.

Add it to your agent workflow

Install the skill with:

npx skills add jeffy-Peng/jeffy-skills --skill 3d-deep-research

Then add a prompt like:

Use 3d-deep-research to analyze the EV battery supply chain. Include X/Y/Z analysis and a full evidence ledger. Deliver Markdown, HTML, and PDF.

The skill reads its reference files, chooses the object adapter, and follows the report template.

Compare before adopting

Compare against other research skills by looking at output quality: does the final report include a claim ledger and an explicit mechanism judgment, or just summaries? Check maintenance: the repo has 20 stars and one primary author, so inspect recent commits before relying on it. Consider alternatives: generic deep-research tools may be better if you only need quick answers without formal traceability. Assess workflow fit: if your agent already produces source-cited reports, this skill adds structure; if you need one-line answers, it is overkill.

Why it is worth tracking

The skill is interesting because it enforces a method rather than adding another “web search then summarize” loop. Its evidence protocol and X/Y/Z framework show real design discipline, and the question-driven chart allocation indicates an author thinking about report quality. With 20 stars it is early, but the pattern is worth tracking for any agent that regularly produces research deliverables. Check its OpenAgentSkill listing at https://www.openagentskill.com/skills/jeffy-peng-3d-deep-research for the current version and adapter coverage.