script-test

name

script-test

description

Generate a companion verification script that exercises a target script with known inputs and asserts expected outputs. Uses the fm-debug infrastructure to report pass/fail results back to the agent. At Tier 1 the developer runs the test script manually. At Tier 3 the agent deploys, runs, and reads results autonomously. Triggers on phrases like "test this script", "write a test", "verification script", "assert results", or "prove this works".

Script Test

Generate a FileMaker test script that exercises a target script with known inputs, captures the results, and reports pass/fail via the Agentic-fm Debug infrastructure.

---

Step 1: Determine the automation tier

Read agent/config/automation.json and check project_tier (preferred) or default_tier.

• Tier 1 — test script goes to clipboard; the developer runs it manually and the agent reads agent/debug/output.json
• Tier 3 — the agent deploys the test script, runs it via the companion, and reads results autonomously

---

Step 2: Understand the target script

Identify the script to test. Read the human-readable version from agent/xml_parsed/scripts_sanitized/ and determine:

1. Parameter format — what does it expect? (JSON, plain text, empty)
2. Exit Script result — what does it return? (JSON, value, empty)
3. Side effects — does it modify records, navigate layouts, create records?
4. Preconditions — what state must exist for it to work? (records in found set, specific layout, specific field values)

If the developer has provided expected inputs and outputs, use those. Otherwise, derive test cases from the script logic.

---

Step 3: Design the test cases

Create a test plan with concrete test cases:

## Test plan: [Target Script Name]

### Test 1: Happy path
- Input: JSONSetElement ( "{}" ; [ "invoiceID" ; "INV-001" ; JSONString ] )
- Expected result: { "success": true }
- Expected side effect: Invoices::Status = "Sent"

### Test 2: Empty parameter
- Input: ""
- Expected result: { "success": false, "error": "missing parameter" }
- Expected side effect: none

### Test 3: Record not found
- Input: JSONSetElement ( "{}" ; [ "invoiceID" ; "NONEXISTENT" ; JSONString ] )
- Expected result: { "success": false, "error": "not found" }
- Expected side effect: none

Present the test plan to the developer for confirmation before generating.

---

Step 4: Generate the test script

Build a test script as fmxmlsnippet that:

1. Sets up preconditions — navigate to the correct layout, ensure test data exists
2. Calls the target script with each test case input via Perform Script
3. Captures the result — Get ( ScriptResult ) immediately after Perform Script
4. Asserts the expected outcome — compares actual vs expected
5. Reports results via Agentic-fm Debug

Test script structure (HR format)

# PURPOSE: Test script for [Target Script Name]
Allow User Abort [ Off ]
Set Error Capture [ On ]

# Build results array
Set Variable [ $results ; "[]" ]
Set Variable [ $testCount ; 0 ]
Set Variable [ $passCount ; 0 ]

# --- Test 1: Happy path ---
Set Variable [ $testName ; "Happy path" ]
Set Variable [ $input ; JSONSetElement ( "{}" ; [ "invoiceID" ; "INV-001" ; JSONString ] ) ]
Set Variable [ $expected ; JSONSetElement ( "{}" ; [ "success" ; True ; JSONBoolean ] ) ]

Perform Script [ "Target Script" ; Parameter: $input ]
Set Variable [ $actual ; Get ( ScriptResult ) ]
Set Variable [ $errData ; JSONSetElement ( "{}" ;
    [ "lastError" ; Get ( LastError ) ; JSONNumber ] ;
    [ "lastErrorDetail" ; Get ( LastErrorDetail ) ; JSONString ] ;
    [ "lastErrorLocation" ; Get ( LastErrorLocation ) ; JSONString ]
) ]

# Assert
Set Variable [ $pass ; JSONGetElement ( $actual ; "success" ) = JSONGetElement ( $expected ; "success" ) ]
Set Variable [ $testCount ; $testCount + 1 ]
Set Variable [ $passCount ; $passCount + If ( $pass ; 1 ; 0 ) ]
Set Variable [ $results ; JSONSetElement ( $results ;
    [ $testCount ; JSONSetElement ( "{}" ;
        [ "test" ; $testName ; JSONString ] ;
        [ "pass" ; $pass ; JSONBoolean ] ;
        [ "expected" ; $expected ; JSONRaw ] ;
        [ "actual" ; $actual ; JSONRaw ] ;
        [ "errData" ; $errData ; JSONRaw ]
    ) ; JSONRaw ]
) ]

# --- (repeat for each test case) ---

# Report results via Agentic-fm Debug
Perform Script [ "Agentic-fm Debug" ; Parameter: JSONSetElement ( "{}" ;
    [ "label" ; "Test results: Target Script" ; JSONString ] ;
    [ "vars"  ; JSONSetElement ( "{}" ;
        [ "testCount" ; $testCount ; JSONNumber ] ;
        [ "passCount" ; $passCount ; JSONNumber ] ;
        [ "allPassed" ; $testCount = $passCount ; JSONBoolean ] ;
        [ "results"   ; $results  ; JSONRaw ]
    ) ; JSONRaw ]
) ]

Exit Script [ JSONSetElement ( "{}" ;
    [ "testCount" ; $testCount ; JSONNumber ] ;
    [ "passCount" ; $passCount ; JSONNumber ] ;
    [ "allPassed" ; $testCount = $passCount ; JSONBoolean ]
) ]

Assertion patterns

Use these comparison patterns depending on what's being tested:

Assertion type	Pattern
Exact match	$actual = $expected
JSON key match	JSONGetElement ( $actual ; "key" ) = "expectedValue"
Not empty	not IsEmpty ( $actual )
Error code	JSONGetElement ( $errData ; "lastError" ) = 0
Contains	Position ( $actual ; "substring" ; 1 ; 1 ) > 0

Look up the Agentic-fm Debug script ID

grep "Agentic-fm Debug" "agent/context/{solution}/scripts.index"

Also look up the target script's ID for the Perform Script step reference.

Validate

python3 agent/scripts/validate_snippet.py agent/sandbox/Test_{ScriptName}.xml

---

Step 5: Deploy and run

Tier 3 (autonomous)

1. Deploy the test script — load clipboard via POST {companion_url}/clipboard, create the script via Tier 3 raw AppleScript (Cmd+N → Rename → Cmd+V → Cmd+S)
2. Run the test — POST {companion_url}/trigger with { "fm_app_name": "...", "script": "Test_{ScriptName}", "target_file": "SolutionName" }
3. Read results — wait 2–3 seconds, then read agent/debug/output.json

Tier 1 (developer-assisted)

Present the test script on the clipboard:

The test script is on your clipboard. To install and run it:

1. In Script Workspace, press Cmd+N to create a new script
2. Rename it to Test_{ScriptName}
3. Cmd+V — paste the test script
4. Cmd+S — save
5. Run the test script
6. Let me know when it's done — I'll read agent/debug/output.json directly.

---

Step 6: Interpret results

Read agent/debug/output.json and parse the test results:

{
  "vars": {
    "testCount": 3,
    "passCount": 2,
    "allPassed": false,
    "results": [
      { "test": "Happy path", "pass": true, ... },
      { "test": "Empty parameter", "pass": true, ... },
      { "test": "Record not found", "pass": false, "expected": "...", "actual": "..." }
    ]
  }
}

For each failing test:

1. Compare expected vs actual to identify the discrepancy
2. Check errData for any FM error that occurred
3. Explain the failure and whether it indicates a bug in the target script or a flaw in the test case

Present a summary:

## Test Results: [Target Script]

✅ 2/3 tests passed

| Test | Result | Notes |
|---|---|---|
| Happy path | PASS | |
| Empty parameter | PASS | |
| Record not found | FAIL | Expected error response, got empty string |

### Failing test analysis
Test 3 failed because the target script exits without a result when
the find returns 0 records. The script should Exit Script with an error
JSON when Get(FoundCount) = 0.

---

Step 7: Clean up

After testing is complete:

• Tier 3: optionally delete the test script from Script Workspace (or leave it for future regression testing — ask the developer)
• Tier 1: inform the developer they can delete the test script if no longer needed

If a bug was found, suggest using script-debug or script-refactor to fix it.

---

Constraints

• Test scripts must not modify production data destructively — if the target script creates or modifies records, the test script should work on test records or restore original values after each test
• Each test must be independent — a failing test should not affect subsequent tests
• Name test scripts with Test_ prefix — e.g., Test_Process Invoice — so they're easy to identify and clean up
• Always include error data capture in the assertion — even passing tests should record $errData so unexpected errors are visible