test-runner

name	test-runner
version	1.0.0
description	Validate generated multi-agent systems: check SKILL.md syntax, verify openclaw.json config, run Lobster workflows against test data, verify result writes to data-ingestion service.
user-invocable	false
metadata	{"openclaw":{"requires":{"bins":["curl","jq"],"env":["DATA_INGESTION_BASE_URL"]}}}

Test Runner Skill

End-to-end validation of generated multi-agent systems.

Test Phases

Phase 1: Static Validation

• Check YAML frontmatter exists and is valid
• Check required fields: name, description, version
• Check metadata.openclaw.requires.env includes DATA_INGESTION_* vars
• Check instructions section is non-empty
• Validate openclaw.json structure
• Validate Lobster workflow DAG

Phase 2: Schema Provisioning

Provision test org/agent schema via data-ingestion service.

Phase 3: Smoke Test Execution

Run each processing skill with mock/minimal input data.

Phase 4: Write Verification (CRITICAL)

Query result_metrics for test run. FAIL if zero rows written.

Phase 5: Cleanup

Delete test data using the data-ingestion delete endpoint.

Output a JSON report with pass/fail status for each phase.