AgentSkillsCN

Test Scenario

测试场景

SKILL.md

Test Scenario Skill

Description

Executes a sandbox scenario to test agent diagnostic capabilities and validates the output against expected diagnosis.

Usage

/test-scenario <scenario-name>

Example: /test-scenario memory-leak

What This Skill Does

  1. •Loads scenario from sandbox/scenarios/<scenario-name>.scenario.ts
  2. •Injects mock tool responses
  3. •Triggers Orchestrator agent with incident report
  4. •Captures agent's diagnostic output
  5. •Validates:
    • •Root cause includes expected keywords
    • •Confidence meets minimum threshold
    • •Remediation suggestions align with expectations
  6. •Reports pass/fail with detailed comparison
  7. •If failed: suggests which agent prompts or tools to adjust

Validation Output Format

code
Scenario: memory-leak
Status: PASSED ✅

Validation Results:
  ✅ Root cause match: Found keywords [OOMKilled, memory, heap]
  ✅ Confidence threshold: 0.89 >= 0.85
  ✅ Remediation alignment: 3/3 suggestions match expected strategies

Agent Diagnosis:
  Root Cause: Memory leak in payment-service causing repeated OOMKills
  Confidence: 0.89
  Evidence:
    - OutOfMemoryError in logs (47 occurrences)
    - Memory usage trending upward to 98% of limit
    - Pod restarted 5 times in last hour
  Remediation:
    - Rollback to v1.5.1 (stable version)
    - Increase memory limit to 2Gi temporarily
    - Profile memory usage to identify leak