Prompt file imported from royosherove/repo-swarm-sample-results-hub (.cursor/commands/report.md). Copyright stays with the author.

Report Generation Rule

Prerequisites

1. Read all *.arch.md files (exclude folders: prompts, diagrams, reports)
2. Apply all cursor rules from .cursor/rules/ folder if it exists, or agents.md or other llm rules folders and files
3. Ignore architecture-hub.arch.md unless explicitly requested

Core Principles

• FACTS ONLY: Every statement must be backed by specific file locations and line references
• No hallucinations: Uncertain information goes in "⚠️ Unverified Items" section
• Name normalization: Treat variants as identical (e.g., llamaindex = llama-index)
• Search in files: search across ALL repos always initially. do not use repo names to decide what should or should not be investigates, since repos can contain things that might not seem related to their name. i.e if you search for "dog", you should not filter for repos whoes names contain dog, but search "dog" inside all arch.md files.
• Multiple search passes might be needed: If you are searching for something that might have depdenencies or two way usages across mutiple servies, you might need to use cli tools multiple passes to search both the producers and then the consumers of something)
• Verification: Double-check for false positives and missing information. note that sometimes in an arch file an item is reported as :searched for", even ift was not found, as part of the disclosure process. so make sure the file mentions it was found, not just searched for. so it should not be counted if it is not specificlaly found.

Execution Process

Phase 1: Planning

1. Create report/.memory/REPORT_PLAN.md
2. Document:
   • Search strategy (IMPORTANT) (might include multiple pases to find collaborating parties such as publishers and subscribers, consumers and producers)
   • Files to analyze
   • Expected output structure
   • CLI tools needed (from your rules or in mise.toml)

Phase 2: Research

1. Execute searches using available CLI tools
2. Extract facts with file references (format: file.md:line)
3. Update REPORT_PLAN.md with progress after each major step
4. Track uncertain items separately

Phase 3: Report Generation

Standard Report Structure:

# [Report Title]

## Executive Summary
[2-3 paragraph overview of key findings]

## Findings Summary
| Item | Category | Location | Quick Summary |
|------|----------|----------|---------------|
| ...  | ...      | ...      | ...           |

[Use sub-groupings if multiple hierarchy levels exist]

## Architecture Diagram 
[Include Mermaid diagram ONLY if relationships exist between items]

## Detailed Findings

### [Item 1]
- **Location**: `path/to/file.md:lines`
- **Details**: [comprehensive details]
- **Related Items**: [if applicable]

[Repeat for each item]

## ⚠️ Unverified Items
[Items requiring clarification, marked with asterisks]

## Reproducibility

### Report Metadata
- **Generated**: [timestamp]
- **Model**: [model name and version]
- **Prompt**: 

[exact prompt used]

- **Context Files**: 
- [list of files analyzed]
- **CLI Tools Used**: 
- [commands executed]

Quality Checklist

Before finalizing, verify:

• All facts have file references
• No false positives included
• No relevant items missed
• Name variants normalized
• architecture-hub.arch.md excluded (unless requested)
• Plan file updated with final status
• Diagram included only if relationships exist
• Diagram should be dark mode compatible
• Diagram in mermaid is the MOST COMPATIBLE so it will render anywhere (avoid errors like "Parse error on line xx: Expecting 'AMP', 'COLON', 'PIPE', 'TESTSTR', 'DOWN', 'DEFAULT', 'NUM', 'COMMA', 'NODE_STRING', 'BRKT', 'MINUS', 'MULT', 'UNICODE_TEXT', got 'LINK_ID'")

Tool Usage

• Leverage CLI tools from cursor rules
• Use mise tools from mise.toml
• Create new reusable tools in mise.toml as needed
• Document tool usage in report metadata