yu-iskw GitHub Agent Testing Skill
Purpose
Automates the verification of GitHub management skills and subagents through 17 practical project management workflows. It distinguishes between Core Agent Scenarios (essential for defined subagents) and Atomic Skill Scenarios (independent or experimental capabilities).
1. Prerequisites
- Target Project: You MUST identify the GitHub project to use for testing.
- Resources: Familiarize yourself with the 17 scenarios and the report template.
2. Initialization
Before starting the tests, you must ask the user which GitHub Project should be the target of the test.
Action: Use the AskQuestion tool to prompt for the Project URL or Number.
3. Test Orchestration Workflow
Follow this checklist to execute the comprehensive test suite:
Group A: Core Agent Scenarios
- Step 0: Identify Target Project (Ask user).
- Step 1: List Open Issues (Baseline).
- Step 2: Create Project-linked Test Bug.
- Step 3: Add Triage Comment.
- Step 4: Apply Categorization Labels.
- Step 5: Remove Temporary Labels.
- Step 6: Assign Lead Maintainer.
- Step 7: Set Release Milestone.
- Step 9: List Global Projects.
- Step 10: Verify Project Linkage.
- Step 11: Retrieve Project Items.
- Step 12: Move Item to 'In Progress'.
- Step 13: Search for Duplicate Issues.
- Step 14: Refine Issue Content (Title/Body).
- Step 15: Multi-Attribute Update (Labels/Milestone).
- Step 16: Move Item to 'Done'.
- Step 17: Close the Verified Issue.
Group B: Atomic Skill Scenarios
- Step 8: Create Linked Branch.
Finalization
- Finalize: Generate Test Report.
4. Instructions
- Capture Timestamps: For every operation, record the ISO timestamp to populate the "Timeline of Operations" in the final report.
- Verify Outcomes: After each atomic skill call, verify that the state in GitHub has changed as expected.
- Handle Errors:
- If a Core (Group A) step fails, it indicates a critical agent regression.
- If an Atomic (Group B) step fails, it may be due to environmental factors (e.g., local git config). Document it as a warning/environmental issue rather than a core failure.
- Reporting: Use the assets/report-template.md to write a detailed Markdown file. Clearly distinguish between Core and Atomic results.
5. Examples
Scenario Execution
Input: "Run the test suite on Project #2." Behavior: The agent lists projects, confirms #2, executes the 17 steps while logging timestamps, and finally outputs the report with categorized results.