gemini-2-5-pro-preview-analysis

name	gemini-2-5-pro-preview-analysis
description	Use Google's gemini-2-5-pro-preview model for deep analysis tasks including document analysis, code review, research synthesis, data interpretation, and multi-modal analysis. Trigger on requests like "analyze this", "deep dive", "research synthesis", "interpret this data".
version	1.0.0
author	Hermes Agent
license	MIT
platforms	["linux","macos","windows"]
metadata	{"hermes":{"tags":["gemini","google","analysis","research","deep-analysis","ai-tools"],"related_skills":["research-paper-writing","llm-wiki","domain-intel"]}}

Gemini 2.5 Pro Preview Analysis

Overview

This skill leverages Google's gemini-2-5-pro-preview model for advanced analysis tasks. The model offers strong reasoning capabilities and extended context window, making it ideal for deep dives into documents, codebases, research papers, and multi-modal content.

gemini-2-5-pro-preview excels at:

• Long-document summarization and extraction
• Multi-step reasoning and analysis
• Code understanding and review
• Cross-document synthesis
• Data interpretation and pattern recognition

When to Use

Use for:

• Deep analysis of long documents (100k+ tokens)
• Multi-document research synthesis
• Codebase architecture analysis
• Technical paper review and interpretation
• Data pattern recognition and insights
• Comparative analysis across sources
• Chain-of-thought reasoning tasks

Don't use for:

• Simple quicklookups (use smaller/faster models)
• Real-time conversations (higher latency than flash models)
• Tasks requiring exact factual recall (prone to confabulation on niche details)

Model Configuration

model: gemini-2-5-pro-preview
provider: google
temperature: 0.0  # Lower for analytical tasks, 0.7-0.9 for creative
max_tokens: 8192
top_p: 0.95

Analysis Patterns

1. Document Deep Dive

When given a document to analyze:

1. First pass: Identify structure, key themes, and scope
2. Second pass: Extract specific insights, claims, evidence
3. Third pass: Synthesize findings into coherent analysis

System: You are a meticulous analyst. Provide thorough, structured analysis with evidence from the source material.

2. Code Review Analysis

For analyzing code:

1. Architecture overview (file structure, dependencies)
2. Logic flow analysis (critical paths, edge cases)
3. Quality assessment (patterns, anti-patterns, security)
4. Improvement recommendations with priority

3. Research Synthesis

For synthesizing multiple sources:

1. Map each source's key contributions
2. Identify consensus and conflicts across sources
3. Highlight gaps, contradictions, or unexplored areas
4. Formulate new insights from combination

4. Data Interpretation

For data analysis tasks:

1. Characterize the data (distribution, types, relationships)
2. Identify patterns, trends, anomalies
3. Statistical significance assessment
4. Interpret findings in context

Output Format

Structure analysis outputs with:

• Executive Summary (2-3 sentences)
• Key Findings (bullet points with evidence)
• Detailed Analysis (expandable sections)
• Implications/Recommendations (if applicable)
• Limitations/Assumptions (honest caveats)

Prompt Templates

Template: Document Analysis

Analyze the following document thoroughly. Identify:
1. Main thesis/argument
2. Supporting evidence and claims
3. Methodology (if applicable)
4. Strengths and weaknesses
5. Implications or conclusions

{Document Content}

Template: Comparative Analysis

Compare and contrast the following sources. Identify:
1. Points of agreement
2. Points of disagreement
3. Unique contributions of each
4. Gaps in coverage
5. Overall synthesis

{Source A}
---
{Source B}

Template: Code Architecture Analysis

Analyze this codebase. Provide:
1. Architecture overview
2. Module dependencies
3. Data flow analysis
4. Design patterns identified
5. Potential issues or improvements

{Code or Repository}

Best Practices

1. Provide context: Give the model background on what kind of analysis you need
2. Set scope: Define boundaries (time period, topic focus, etc.)
3. Ask for evidence: Request citations/quotes from source material
4. Iterate: Start broad, then drill into specific aspects
5. Verify: Cross-reference key claims, especially for factual content

Common Pitfalls

Pitfall	Prevention
Vague requests	Specify output format and focus areas upfront
Missing context	Provide background and scope definition
Over-reliance on single source	Cross-verify with multiple sources
Ignoring model limitations	Flag uncertainty on niche factual claims
Scope creep	Break large analysis into multiple focused passes

Verification Checklist

• Clear analytical framework applied
• Evidence cited from source material
• Key findings summarized at top
• Limitations acknowledged
• Actionable insights provided (if applicable)