tf-e2e-tester

name	tf-e2e-tester
description	Non-interactive test harness for end-to-end Terraform workflow testing. Runs full `/tf-plan` → `/tf-implement` cycle with test defaults, bypassing user prompts for automated validation.
args	<prompt-file>

E2E Test Orchestrator

Workflow Steps

• Execute the phases sequentially. Do not move ahead until the step previous has completed.
• Before and after each subagent run: post progress to the gh issue. After each subagent run: commit its output to git.
• Use the following template for gh issues:

bash .foundations/scripts/bash/post-issue-progress.sh $ISSUE_NUMBER "<step>" "<status>" "<summary>" "$DETAILS"

example

bash .foundations/scripts/bash/post-issue-progress.sh $ISSUE_NUMBER "Research" "started"
bash .foundations/scripts/bash/post-issue-progress.sh $ISSUE_NUMBER "Research" "complete" $DETAILS

E2E Overrides

These overrides replace interactive prompts with test defaults:

Override	Behavior
Requirements	Read from .claude/skills/tf-e2e-tester/prompts/<prompt-file>
AskUserQuestion	Use test defaults, do not prompt
Approval gates	Auto-approve, do not wait
Destroy sandbox	Always yes
Create PR	No, test artifacts stay on branch

---

PART 1: PLANNING (from /tf-plan)

Phase 1 — Setup

• Run validate-env.sh
• Check for error and stop if pre-requisites not met
• Then call MCP list_terraform_orgs to verify TFE_TOKEN.
• E2E Override: Read requirements from .claude/skills/tf-e2e-tester/prompts/<prompt-file> instead of gathering from user
• Create test issue with test:e2e label:

  gh issue create --title "E2E Test: <prompt-file>" --label "test:e2e" --body "$(cat .claude/skills/tf-e2e-tester/prompts/<prompt-file>)"
  

Phase 2 — Specification (sequential)

1. sdd-specify → spec.md
2. sdd-checklist against spec.md → checklist findings
3. sdd-clarify with checklist findings + spec.md. E2E Override: Use test defaults for HIGH-impact gaps, do not use AskUserQuestion

Phase 3 — Research + Planning (sequential)

1. sdd-research (parallel) → research findings
2. sdd-plan-draft → plan.md, data-model.md, contracts/module-interfaces.md

Phase 4 — Tasks + Analysis (sequential)

1. sdd-tasks → tasks.md
2. sdd-analyze with tasks.md → analysis.md

Phase 5 — Summary (no approval wait)

• Compile results, post to GitHub issue
• E2E Override: Do NOT add agent:awaiting-review label. Do NOT stop for approval. Proceed directly to implementation.

---

PART 2: IMPLEMENTATION (from /tf-implement)

Phase 1 — Prerequisites

• Use issue number from Part 1
• Verify artifacts exist in $FEATURE_DIR: spec.md, plan.md, tasks.md, contracts/module-interfaces.md

Phase 2 — Implementation (sequential by phase)

For each phase in tasks.md:

1. tf-task-executor with phase tasks + plan.md + contracts/module-interfaces.md
2. Mark completed tasks [X] in tasks.md
3. Commit: test(e2e): implement phase N - <description>

Phase 3 — Design Review

• Run aws-security-advisor and code-quality-judge subagents in parallel against plan artifacts

Phase 4 — Sandbox Testing

• Run tf-deployer subagent
• Capture run URL and deploy status

Phase 5 — Report

• Run tf-report-generator subagent with run URL and deploy status
• Gate: report must exist at specs/<branch>/reports/deployment_*.md

Phase 6 — Cleanup

• git push (do NOT create PR)
• Post completion comment to issue
• E2E Override: Always destroy sandbox resources via tf-deployer
• Close issue with test:passed or test:failed label

Display: > E2E test complete. Status: [PASSED|FAILED]. See issue # for details.