User Input

$ARGUMENTS

• Read Issue Template: Read .github/ISSUE_TEMPLATE/terraform-agent-provisioning.yml to understand required fields

• Gather User Inputs: Use user's natural language request to populate issue fields wherever possible

• Create GitHub Issue:

• Use gh issue create with the user provided input and template values

• Title format: [AGENT PROVISION] <descriptive-name>

• Labels: agent-driven , terraform , infrastructure , provisioning

• Populate all fields from the template

• Validate Issue: Confirm the GitHub issue is valid and contains all required information

• Mark as In Progress: Add in-progress label when starting work using gh issue edit <issue-number> --add-label "in-progress"

• Update Issue with Progress: Comment on the issue at the start and completion of each Github spec-kit stage with a short summary and link to the generated artifacts:

• Format: 🤖 [Stage Name] - [Started/Completed]: Brief summary

• Example: 🤖 speckit.specify - Started: Creating feature specification from requirements

• Example: 🤖 speckit.specify - Completed: Generated spec.md with 5 core requirements

Environment

All files and folders exist in /workspace/ directory. All speckit scripts are located in /workspace/.specify/scripts/bash directory.

You should not require to change to any other directory.

Execution Workflow

For each task use concurrent subagents to speed up the process. At each stage, commit changes and update the GitHub issue with progress comments. Allocate subtask efficiently, some tasks like writing a file should only be created once, I will breakdown tasks to you.

• Create and configure tracking GitHub issue from template. Github issue should be created and labeled appropriately. Confirm the gh issue is valid, when you start mark the issue to in-progress using the label in-progress, update the github issue with comments when you start and finish each speckit stage with a short summary

• Validate environment and credentials by running .specify/scripts/bash/validate-env.sh

• Use concurrent subagent speckit.specify

• Create feature specification from the issue details and continue to next stage. Output Github issue number and branch to gh-issue.json for use by implementation agent

• commit and update Git issue and continue to next stage

• Use concurrent subagent speckit.clarify and continue to next stage

• commit and update Git issue and continue to next stage

• Use concurrent subagent speckit.plan and continue to next stage

• commit and update Git issue and continue to next stage

• Use concurrent subagent speckit.tasks and continue to next stage

• commit and update Git issue and continue to next stage

• Request user to review and approve design (human-in-the-loop) before implementation phase

• Update Git issue with user details and approval status.

• Ensure that gh-issue.json contains all necessary details for implementation agent to proceed.

GitHub Issue Template Mapping

When creating the issue, map user inputs to these key template fields:

Required Fields:

• hcp_org : HCP Terraform organization name

• hcp_project : HCP Terraform project name

• workspace_name : Workspace name (use pattern: sandbox_<REPO_NAME> for testing)

• terraform_version : Terraform version (default: "Latest (recommended)")

• project_name : Project/application name

• cloud_provider : AWS, Azure, GCP, Multi-cloud, or Other

• cloud_region : Primary cloud region

• environment : development, staging, production, sandbox, test, or dr

• infrastructure_components : Detailed list of components to provision

Optional but Important:

• additional_regions : Multi-region deployments

• existing_infrastructure : Existing resources to reference

• module_preference : "Private Registry Only (recommended)" is default

• security_requirements : Security controls checklist

• configuration_values : Key configuration parameters

• network_requirements : Network features needed

• agent_autonomy : Level of autonomy (default: "Fully Autonomous")

Agent Instructions

When user provides infrastructure request:

• Extract all available information from natural language input

• Read the issue template to understand all required and optional fields

• Map user's request to template fields (infer reasonable defaults where needed)

• Create GitHub issue with gh issue create using extracted values

• Validate the created issue has all critical information

• Add in-progress label before starting work

• Post progress comments at start/completion of each Speckit phase

• Request user to review and approve design (human-in-the-loop) before implementation phase