Xlsx testcases

/xlsx-testcases — bridge tester xlsx ↔ runnable tests

Keeps xlsx as source of truth for test cases (testers' familiar tool), generates runnable artifacts on demand, and syncs to Azure DevOps Test Plans.

Deps: anthropic-skills:xlsx (delegate parsing), jq, python3. ADO sync uses /azure-devops skill (multi-profile).

Overview

Bridges xlsx-based test case authoring (testers' familiar tool) with runnable test specs. Generates Maestro YAML / Detox / Markdown on demand and syncs to Azure DevOps Test Plans. xlsx remains the source of truth — generated artifacts are derivable.

When to Use

• QA team maintains test cases in xlsx (no migration possible)
• Need runnable artifacts (Maestro YAML, Detox, Markdown checklists)
• Two-way sync with Azure DevOps Test Plans
• Coverage gap reports (TCs in xlsx not yet automated)

When NOT to Use

• Test cases live in code already → no need for this bridge
• One-off conversion → use anthropic-skills:xlsx directly
• Real-time test execution → this is scaffolding, not a runner
• Non-tabular test specs (BDD .feature files) → use Cucumber / Gherkin tooling

Schema mapping

Per-workbook schema at <xlsx-folder>/.testcase-schema.yml (committed):

sheet: "Test Cases"           # or "*" for all sheets
header_row: 2
columns:
  id:         "TC ID"          # required, unique
  module:     "Module"
  title:      "Title"          # required
  steps:      "Steps"          # multi-line cell
  expected:   "Expected Result"
  priority:   "Priority"
  category:   "Category"       # smoke/regression/etc.
  status:     "Status"         # New/Updated/Stable
parsing:
  steps_separator: "newline"   # or "numbered"
  priority_map:
    P0: high
    P1: high
    P2: medium
    P3: low

If no schema file exists, prompt to infer from first sheet header row and write file for review.

Dispatch

parse <xlsx-path> → /tmp/xlsx-testcases-<hash>.json

Delegates to anthropic-skills:xlsx for parsing. Cache key = SHA-256 of xlsx mtime+size+schema content. Output array of testcase objects.

gen maestro <xlsx-path> [--out tests/maestro/] — Maestro YAML

For each TC: tests/maestro/<id>-<slug>.yaml. Step translation heuristic:

xlsx step pattern	Maestro action
"Open app" / "Launch"	- launchApp
"Tap [X]"	- tapOn: "X"
"Enter [X] in [field]"	- tapOn: "field" + - inputText: "X"
"Verify [X] visible"	- assertVisible: "X"

Unmapped steps emit # UNMAPPED: comment and surface in run summary.

gen detox <xlsx-path> [--out e2e/] — Detox .test.ts

Same flow as Maestro but JavaScript spec.

gen markdown <xlsx-path> — manual QA checklist

## TC001 — Title  [tags, priority]
**Precondition:** ...
**Steps:** 1. ... 2. ...
**Expected:** ...
[ ] Pass  [ ] Fail  [ ] Skip   Notes: ___

sync ado <xlsx-path> [--plan <id>] [--dry-run]

Sync to Azure DevOps Test Plans (uses /azure-devops skill auth):

1. Search existing test cases by TC ID
2. Update if found (PATCH), else create (POST)
3. Track sync state in <xlsx-folder>/.testcase-sync-state.json (committed) — only sync TCs whose hash changed since last sync.

PAT scope required: Test Plans (Read & Write) + Work Items (Read & Write).

--dry-run shows what would change without writing.

coverage <xlsx-path> --against <test-folder> — gap report

Total TCs in xlsx:       142
Generated as Maestro:    98 (69%)
Synced to ADO:           120 (85%)

HIGH-priority untested TCs:
  TC047  IAP   Restore purchase across devices
  TC089  Push  Background notification handling

diff <xlsx-path> — what changed since last sync

Compare current xlsx against .testcase-sync-state.json:

Added: 3 TCs    Modified: 12 TCs    Removed: 1 TC

Implementation notes

• Delegate parsing to anthropic-skills:xlsx, don't reinvent.
• Step translation is heuristic — emit UNMAPPED comments, never silently drop.
• Cache key from input hash → no re-parse on repeated invocation.
• Per-TC sha256(title+steps+expected+priority) → ADO sync only delta.
• UTF-8 throughout (Vietnamese / Asian content must round-trip).

Safety

• xlsx may contain sensitive test data (sample passwords, PII). Treat as confidential: cache to /tmp/, don't include verbatim in chat beyond what user asked.
• ADO sync is mutating → always run --dry-run first; require ctt_confirm before actual sync. Audit log records TC IDs only, not content.
• Generated runners are committed code — review before pushing public.
• .testcase-schema.yml and .testcase-sync-state.json are safe to commit (no test data, only mapping + hashes).

Common Mistakes

• No .testcase-schema.yml committed → schema inference per-run is non-deterministic
• Step text "Tap submit" mapped to tapOn:"submit" then UI uses different label → test fails. Review unmapped comments.
• Skipping --dry-run before ADO sync → hard to undo bulk changes
• xlsx file in git as binary → diffs unreadable. Commit alongside .testcase-schema.yml at minimum.
• UTF-8 encoding lost during sync → Vietnamese / Asian characters mangled
• Editing generated YAML without source xlsx update → drift; xlsx is source of truth

Workflow for QA team

1. Tester writes test cases in xlsx (current habit, no change).
2. Lead/Dev runs /xlsx-testcases gen maestro <xlsx> to scaffold runners.
3. Dev edits scaffolded YAML to add real selectors / data refs.
4. CI runs Maestro on every PR.
5. Weekly: lead runs sync ado + coverage to identify gaps.