Instruction file imported from qprjack86/azure-agentic-infraops-base (.github/instructions/agent-research-first.instructions.md). Copyright stays with the author.

Agent Research Requirements

MANDATORY: All agents MUST perform thorough research before implementation.

This instruction enforces a "research-first" pattern to ensure complete, one-shot execution without missing context or requiring multiple iterations.

Pre-Implementation Research Checklist

Before creating ANY output files or making changes, agents MUST:

• Search workspace for existing patterns (agent-output/, similar projects, templates)
• Read relevant templates in .github/skills/azure-artifacts/templates/ for output structure
• Query documentation via MCP tools (Azure docs, best practices)
• Validate inputs - confirm all required artifacts from previous steps exist
• Check shared defaults in .github/skills/azure-defaults/SKILL.md
• Achieve 80% confidence before proceeding to implementation

Research Workflow Pattern

<research_mandate> MANDATORY: Before producing output artifacts, run comprehensive research.

Step 1: Context Gathering (REQUIRED)

Use read-only tools to gather context without making changes:

# Workspace context
- semantic_search: Find related code, patterns, and documentation
- grep_search: Search for specific terms, resource names, patterns
- read_file: Read templates, existing artifacts, configuration files
- list_dir: Explore project structure

# Azure context (where applicable)
- Azure MCP tools: Query documentation, best practices, SKU info
- mcp_bicep_list_avm_metadata: Check Azure Verified Module availability

Step 2: Validation Gate (REQUIRED)

Before implementation, confirm:

1. Required inputs exist - Previous step artifacts are present and complete
2. Templates loaded - Output structure template has been read
3. Standards understood - Shared defaults and naming conventions reviewed
4. Azure guidance obtained - Relevant documentation queried

Step 3: Confidence Assessment

Only proceed when you have 80% confidence in your understanding of:

• What needs to be created/modified
• Where files should be located
• What format/structure to use
• What Azure resources/patterns apply

If confidence is below 80%: Use #tool:agent to delegate autonomous research, or ASK the user for clarification rather than assuming. </research_mandate>

Delegation Pattern for Deep Research

When extensive research is needed, delegate to a subagent:

MANDATORY: Run #tool:agent tool, instructing the agent to work autonomously
without pausing for user feedback, to gather comprehensive context and return findings.

This pattern enables thorough investigation without interrupting the workflow.

Enforcement Rules

DO:

• Research BEFORE creating files
• Read templates BEFORE generating output
• Query Azure docs BEFORE recommending services
• Check existing patterns BEFORE creating new ones
• Validate inputs BEFORE proceeding to next step
• Ask for clarification when confidence is low

DO NOT:

• Skip research to "save time"
• Assume requirements without verification
• Create output without reading the template first
• Recommend Azure services without checking documentation
• Proceed with missing inputs from previous workflow steps
• Make up information when uncertain

Per-Agent Research Focus

Integration with Workflow

This research-first pattern integrates with the multi-step workflow:

1. Each step should validate outputs from previous steps exist
2. Each step should read its output template before generating content
3. Each step should query relevant Azure documentation
4. Each step should achieve 80% confidence before proceeding

See Azure Defaults Skill for the complete research requirements specification.