agentdevsl/tf-consumer-plan

SDD — Consumer Plan

Produces specs/{FEATURE}/consumer-design.md from requirements. Stops for human approval before any code is written.

Post progress at key steps: bash .foundations/scripts/bash/post-issue-progress.sh $ISSUE_NUMBER "<step>" "<status>" "<summary>". Valid status values: started, in-progress, complete, failed. Checkpoint after each phase: bash .foundations/scripts/bash/checkpoint-commit.sh "<step_name>". The <step_name> must be a short hyphenated identifier (e.g., "clarify", "research-and-design", "design-approved") — NOT a sentence or file path.

Phase 1: Requirements & Research

1. Run bash .foundations/scripts/bash/validate-env.sh --json. Stop if gate_passed=false. Then call MCP list_terraform_orgs to verify TFE_TOKEN — consumer workflows deploy to HCP Terraform, so this is critical.
2. Parse $ARGUMENTS for project name, provider, and description. Ask via AskUserQuestion if incomplete.
3. Create GitHub issue: read .foundations/templates/issue-body-template.md, fill in the placeholders with parsed requirements (adapt for consumer context — modules composed, not resources created), and run gh issue create --title "Consumer: {project-name}" --body "$FILLED_BODY". Capture $ISSUE_NUMBER. Update the issue body again after Step 6 (clarification) to include module selections and scope boundaries.
4. Create feature branch: bash .foundations/scripts/bash/create-new-feature.sh --json --workflow consumer --issue $ISSUE_NUMBER --short-name "<project-name>" "<feature description>". Parse the JSON output to capture $BRANCH_NAME as $FEATURE and $DESIGN_FILE.
5. Scan requirements against the tf-domain-category skill — focus on module composition ambiguity, networking integration, and workspace configuration decisions.
6. Ask up to 4 clarification questions via AskUserQuestion. Must include:
   • Module selection: Which private registry modules are required and their approximate versions?
   • Environment and workspace: Target workspace, region, and credential pattern (dynamic credentials, assume_role)?
   • A security-related question (encryption, public access, IAM)
   • Scope/integration clarification as needed (networking, monitoring, cross-workspace dependencies)
7. Launch 3-4 concurrent tf-consumer-research subagents (run in foreground — they use MCP tools):
   • Private registry modules: Available modules, versions, inputs, outputs, secure defaults
   • Module wiring: How module outputs connect to inputs, type compatibility, composition patterns
   • Workspace and deployment: HCP Terraform workspace configuration, variable sets, dynamic credentials Wait for all to complete. Verify research files exist at specs/{FEATURE}/research-*.md via Glob.

Phase 2: Design

8. Launch tf-consumer-design agent with FEATURE path and clarified requirements. The agent reads the constitution, design template, and research files from specs/{FEATURE}/research-*.md itself.
9. Verify specs/{FEATURE}/consumer-design.md exists via Glob. Re-launch once if missing.
10. Grep to confirm all 6 sections present (## 1. Purpose through ## 6. Open Questions). Fix inline if any missing.
11. Present design summary to user via AskUserQuestion: module count, wiring connection count, variable count, security controls, checklist items. Options: approve, review file first, request changes.
12. If changes requested, apply and re-present. Repeat until approved.

Done

Design approved at specs/{FEATURE}/consumer-design.md. Run /tf-consumer-implement $FEATURE to build.