Looplia E2E Test Skill

Comprehensive end-to-end testing for looplia CLI covering Docker deployments, published CLI validation, and local development testing.

Quick Reference

Test Modes

Mode 1: Docker E2E (Primary)

Build and test looplia in a Docker container. Best for CI and reproducible testing.

# Set API key
export ZENMUX_API_KEY=xxx

# Run Docker E2E test
./scripts/docker-e2e.sh

What it tests:

• Docker image builds successfully
• CLI initializes workspace
• Provider configuration works
• Workflow executes and produces outputs
• v0.6.9+ subagent architecture

See scripts/docker-e2e.sh for full implementation.

Mode 2: Published CLI Testing

Test the published @looplia/looplia-cli package after CI passes.

# Set API key
export ZENMUX_API_KEY=xxx

# Test latest published version
./scripts/published-cli-e2e.sh

# Test specific version
./scripts/published-cli-e2e.sh 0.6.10

What it tests:

• Published package installs correctly
• Fresh workspace bootstrap works
• Workflow execution with real package

Mode 3: Local Development

Quick testing during development.

# Build and run locally
bun run build
bun run dev build "test prompt"

# Or use installed CLI
looplia --version
looplia build --mock "test prompt"

Mode 4: v0.6.10 Verification

Specific checks for v0.6.10 unified command initialization.

./scripts/check-v0610.sh

What it tests:

• Mock mode works without API key
• ZenMux API key mapping
• Error message format
• Settings loading order

Common Verification Functions

Source scripts/verify-workflow.sh to access common verification functions:

source scripts/verify-workflow.sh

# Find sandbox directory
SANDBOX=$(find ~/.looplia/sandbox -maxdepth 1 -type d ! -name sandbox | head -1)

# Run verifications
verify_outputs "$SANDBOX"
verify_validation_state "$SANDBOX"
verify_subagent_usage "$SANDBOX"
verify_v0610_init

Expected Outputs

A successful workflow run produces:

~/.looplia/sandbox/<run-id>/
├── outputs/
│   ├── summary.json      # Stage 1: Content analysis
│   ├── ideas.json        # Stage 2: Idea generation
│   └── writing-kit.json  # Stage 3: Final output
├── validation.json       # All steps validated: true
└── logs/
    └── *.log             # Execution logs

v0.6.9+ Architecture Verification

The v0.6.9 architecture uses general-purpose subagent for all workflow steps.

Expected in logs:

• "subagent_type": "general-purpose" appears >= 3 times
• No legacy agents: content-analyzer, idea-generator, writing-kit-builder

Check with:

grep -c '"subagent_type".*"general-purpose"' ~/.looplia/sandbox/*/logs/*.log
# Expected: >= 3

v0.6.10 Command Init Verification

The v0.6.10 architecture unifies command initialization.

Key tests:

1. looplia build --mock "test" - Works without API key
2. looplia build with ZENMUX_API_KEY - No longer fails with "API key required"
3. Error messages show all options

See references/VERIFICATION.md for detailed verification steps.

Success Criteria

• CLI installed and version displayed
• Workspace bootstrapped with 2 plugins (looplia-core, looplia-writer)
• Provider configured (ZenMux or Anthropic)
• Workflow completed successfully
• All 3 output files created (summary.json, ideas.json, writing-kit.json)
• writing-kit.json passes schema validation
• All steps show validated: true in validation.json
• Logs show general-purpose subagent usage (v0.6.9+)
• No legacy agents detected in logs
• Hook count >= 2, outline sections >= 3

Troubleshooting

See references/TROUBLESHOOTING.md for common issues and solutions.

Quick fixes:

• API key issues: export ZENMUX_API_KEY=xxx
• Docker issues: docker info to verify Docker is running
• Workspace issues: rm -rf ~/.looplia && looplia init --yes

File Structure

.claude/skills/looplia-e2e/
├── SKILL.md                    # This file
├── scripts/
│   ├── docker-e2e.sh           # Docker E2E test
│   ├── debug-docker-e2e.sh     # Debug Docker E2E with ZenMux
│   ├── published-cli-e2e.sh    # Published CLI test
│   ├── verify-workflow.sh      # Common verification functions
│   └── check-v0610.sh          # v0.6.10 specific checks
├── references/
│   ├── VERIFICATION.md         # Detailed verification guide
│   └── TROUBLESHOOTING.md      # Error handling guide
└── assets/
    └── ai-healthcare.md        # Test content fixture

Environment Variables

Cost Optimization: Always use ZENMUX_API_KEY for E2E testing. The ZenMux GLM 4.7 preset is significantly cheaper than Anthropic direct API.

*One of ZENMUX_API_KEY or ANTHROPIC_API_KEY is required (unless using --mock).

See Also

• references/VERIFICATION.md - Success criteria and validation
• references/TROUBLESHOOTING.md - Common issues and solutions
• docs/DESIGN-0.6.9.md - Subagent architecture design
• docs/DESIGN-0.6.10.md - Unified command initialization design