digraph flow {
  rankdir=TB;
  "Prompt for Linear ticket" -> "Pull latest main";
  "Pull latest main" -> "Read ticket via Linear MCP";
  "Read ticket via Linear MCP" -> "Classify issue type";
  "Classify issue type" -> "Plan exploration strategy";
  "Plan exploration strategy" -> "User approves plan?";
  "User approves plan?" -> "Dispatch explore agents" [label="yes"];
  "User approves plan?" -> "Plan exploration strategy" [label="revise"];
  "Dispatch explore agents" -> "Search agno GitHub for duplicates";
  "Search agno GitHub for duplicates" -> "Synthesize and write markdown";
  "Synthesize and write markdown" -> "Format with prettier";
}

Step 1: Collect Context

Prompt the user for the Linear ticket (ID like ENG-123 or a full Linear URL).

Pull latest changes before any research:

git checkout main && git pull origin main

Read the ticket using these MCP tools:

mcp__linear-server__get_issue with includeRelations: true
mcp__linear-server__list_comments for discussion context

Step 2: Classify

Determine issue type from ticket content:

Type	Signals
Bug Report	Error reports, unexpected behavior, regressions
Feature Request	Enhancements, missing capabilities, improvements
Question / Discussion	Clarifications, architectural or usage questions

Confirm classification with the user before proceeding.

Step 3: Plan Exploration Strategy

This is the most critical step. Spend significant time here.

Before dispatching any agents:

Read the Linear ticket thoroughly — understand what's being asked
Identify which agno modules/directories are relevant (agents, tools, workflows, models, knowledge, teams, etc.)
Formulate specific research questions that need answering
Determine how many explore agents are needed — scale to complexity, don't over-dispatch
Assign each agent a clear, non-overlapping research scope

Present the exploration plan to the user for approval before dispatching.

Step 4: Research

Dispatch explore agents using the Task tool with subagent_type: "Explore". Run agents in parallel where scopes are independent.

Each agent should return:

Relevant file paths with line numbers
Code patterns and implementations found
Related existing behavior
Potential impact areas

Step 5: Check for Duplicates

Search the agno GitHub repo for related issues:

Use mcp__github__search_issues with q: "<keywords> repo:agno-agi/agno"
Note any related open or closed issues

Step 6: Write Output

Write to specs/issue-{ticket-number}.md using the appropriate template.

Bug Report

# [Title]

## Description

## Steps to Reproduce
1.

## Expected Behavior

## Actual Behavior

## Environment
- OS:
- Python version:
- Agno version:

## Code References

## Related Issues

Feature Request

# [Title]

## Problem Description

## Proposed Solution

## Additional Context

## Related Issues

Question / Discussion

# [Title]

## Question

## Relevant Code

## Related Issues

Step 7: Format with Prettier

After writing the markdown file, format it with prettier:

npx prettier --write specs/issue-{ticket-number}.md

Display the full content to the user for review after formatting.

agno-issue-explorer

Agno Issue Explorer

Workflow