Agent Skills: Iterative Analysis

Performs sequential multi-round oracle analysis where each round builds on previous findings. Use when user specifies analysis count (e.g., "analyze 5 times"), asks to think repeatedly, or needs progressive refinement.

UncategorizedID: meglinge/megskills/iterative-analysis

Install this agent skill to your local

pnpm dlx add-skill https://github.com/meglinge/megskills/tree/HEAD/skills/iterative-analysis

Skill Files

Browse the full folder contents for iterative-analysis.

Download Skill

Loading file tree…

skills/iterative-analysis/SKILL.md

Skill Metadata

Name
iterative-analysis
Description
Performs sequential multi-round oracle analysis where each round builds on previous findings. Use when user specifies analysis count (e.g., "analyze 5 times"), asks to think repeatedly, or needs progressive refinement.

Iterative Analysis

Performs multi-round, progressively deeper analysis using the oracle. Each iteration builds upon previous findings to refine understanding.

When to Use

  • User asks for "deep thinking" or "thorough analysis"
  • Complex problems requiring multiple perspectives
  • When initial analysis needs refinement
  • User explicitly requests N rounds of analysis

Parameters

  • Iterations: Number of analysis rounds (default: 3, user can specify)
  • Focus: The topic or problem to analyze
  • Files: Relevant files to examine

Workflow

Step 1: Get User Requirements

Ask user (if not specified):

  • What to analyze
  • How many iterations (default: 3)
  • Any specific focus areas

Step 2: Execute Iterative Analysis

For each iteration (1 to N):

Round 1 - Initial Analysis:
oracle(
  task: "Analyze [topic]. Provide initial findings, identify key areas needing deeper investigation.",
  files: [relevant_files]
)

Round 2 - Deeper Investigation:
oracle(
  task: "Based on previous findings: [summary of round 1]. 
         Dig deeper into identified issues. Challenge initial assumptions. 
         Find edge cases and overlooked aspects.",
  files: [relevant_files],
  context: "[Round 1 findings]"
)

Round 3 - Refinement & Synthesis:
oracle(
  task: "Based on all previous analysis: [summary of rounds 1-2].
         Synthesize findings. Identify patterns. Provide final recommendations.
         Rate confidence level for each conclusion.",
  files: [relevant_files],
  context: "[All previous findings]"
)

... continue for additional rounds as specified

Step 3: Final Report

After all iterations, compile:

  • Key discoveries from each round
  • Evolution of understanding
  • Final conclusions with confidence levels
  • Actionable recommendations

Iteration Prompts Template

| Round | Focus | |-------|-------| | 1 | Initial exploration, identify key areas | | 2 | Challenge assumptions, find edge cases | | 3 | Cross-validate, identify patterns | | 4 | Stress-test conclusions, find gaps | | 5+ | Synthesize, prioritize, finalize |

Example Usage

User: "帮我分析这个认证系统,进行5轮分析"

Round 1: "Analyze the authentication system architecture and flow"
Round 2: "Based on [R1], investigate security vulnerabilities and edge cases"
Round 3: "Based on [R1,R2], analyze error handling and failure modes"
Round 4: "Based on [R1-R3], evaluate scalability and performance implications"
Round 5: "Synthesize all findings, prioritize issues, provide remediation plan"

Best Practices

  1. Carry Context Forward: Each round must reference previous findings
  2. Progressive Focus: Narrow focus as iterations progress
  3. Challenge Assumptions: Later rounds should question earlier conclusions
  4. Track Evolution: Note how understanding changed across rounds
  5. Confidence Levels: Final round should rate certainty of each finding