{
  "skill": "architecture-decision-record",
  "version": "1.0.0",
  "testCases": [
    {
      "id": "basic-we-need-to-decide",
      "rule": "",
      "query": "We need to decide between PostgreSQL and MongoDB for our user service. Document this as an ADR.",
      "expectedBehavior": [
        "Creates ADR following Nygard template with numbered title (ADR-NNNN)",
        "Includes Context section with problem statement and constraints",
        "Documents Decision with specific actionable language",
        "Lists Consequences categorized as positive, negative, and neutral outcomes",
        "Documents at least 2 Alternatives Considered with rationale for rejection",
        "Sets Status to Proposed indicating the decision awaits approval"
      ]
    },
    {
      "id": "negative-we-previously-decided-on",
      "rule": "",
      "query": "We previously decided on REST (ADR-0003) but now want to move to GraphQL. Create an ADR that supersedes the old one.",
      "expectedBehavior": [
        "Creates new ADR with next sequential number, does NOT edit ADR-0003",
        "Sets Status to Proposed with reference to superseded ADR-0003",
        "Explains why the original REST decision is being revisited",
        "Documents the evolution context and new forces at play",
        "Links back to ADR-0003 in References section"
      ]
    },
    {
      "id": "negative-fix-the-bug-in",
      "rule": "",
      "query": "Fix the bug in our login endpoint that returns 500 when email is null",
      "expectedBehavior": [
        "Claude does NOT invoke the architecture-decision-record skill",
        "Treats this as a bug fix, not an architectural decision",
        "Uses standard debugging and code editing tools instead",
        "Does not create an ADR document"
      ]
    },
    {
      "id": "interrogation-reliability",
      "rule": "interrogation-reliability",
      "query": "We're deciding on the data architecture for our user service. What reliability questions should we ask?",
      "expectedBehavior": [
        "Asks where data naturally belongs and what the primary access pattern is for the service",
        "Questions retention policy and flags 'keep everything' as a red flag answer",
        "Evaluates UX impact including what happens when the service is down or degraded",
        "Flags vague answers like 'both reads and writes' as insufficient for architectural decisions"
      ]
    },
    {
      "id": "interrogation-scalability",
      "rule": "interrogation-scalability",
      "query": "Before committing to this architecture, what scalability questions should we answer?",
      "expectedBehavior": [
        "Asks about expected user count, data volume now and in one year, and request rate",
        "Questions whether complexity grows linearly or exponentially with scale increase",
        "Evaluates behavior at 10x and 100x current load to identify scaling bottlenecks early",
        "Flags 'it won't be a problem' as a red flag that scale analysis has not been done"
      ]
    },
    {
      "id": "interrogation-security",
      "rule": "interrogation-security",
      "query": "What security questions should we ask before finalizing this architecture decision?",
      "expectedBehavior": [
        "Asks who can access the data and how tenant isolation is enforced at the architecture level",
        "Questions what attack vectors the decision introduces and what happens if authorization fails",
        "Flags 'we trust the frontend' for tenant isolation as a critical security red flag",
        "Evaluates whether PII is involved and what handling requirements apply to the data"
      ]
    }
  ]
}