/sap-test-replay

name: sap-test-replay
description: |
  Turns a recorded business scenario into a re-runnable pass/fail GUI regression test — the
  eCATT/CBTA equivalent the suite lacks — so a regression cycle stops meaning "re-test transactions
  by hand and paste screenshots into Word". A scaffold-recorded linear scenario (control IDs + screen
  identity + popup transitions from /sap-gui-probe) is compiled into GUI segments run by ONE generic
  interpreter VBS, with exactly three checkpoint types — field (control value), message (status-bar
  class/number/type, can capture the created doc number into a token), and table (post-step
  RFC_READ_TABLE key assertion) — and a screenshot on every failure. The headline value AI adds is
  honest triage: FAIL (the system regressed) is kept STRICTLY distinct from REPLAY_ERROR (the replay
  broke: guard mismatch / unexpected popup / capture failure), so flakiness never masquerades as a
  regression; a table check that can't run is COULD_NOT_CHECK -> PASS_WITH_GAPS, never a silent pass.
  run executes transactions (confirm-gated; typed confirm off DEV); lint validates token coverage +
  guard completeness + tcode + checkpoint fields vs live DDIC; init converts a probe folder into a
  scenario skeleton. Emits a sapdev.testverdict/1 line so a campaign ledger and /sap-run-abap-unit
  join the same verdict stream. Release variance lives in scenario DATA, not VBS code. Prerequisites:
  a scaffold-recorded scenario (/sap-gui-probe + init); /sap-login session; NCo 3.1 (32-bit).
argument-hint: "run <scenario.replay.json> [--data <bindings.tsv>] [--keep-going] [--dry-run] | lint <scenario.replay.json> [--data <b.tsv>] | init <probe-folder> [--name <s>]"

SAP Test-Replay Skill

You replay a recorded scenario as a pass/fail regression test: compile it, lint it, drive the GUI via the generic interpreter, assert checkpoints (field / message / RFC table), and roll up a verdict that keeps FAIL (regression) strictly apart from REPLAY_ERROR (the replay broke). run is confirm-gated — it executes real transactions.

Task: $ARGUMENTS

The compile/lint, table checks, and roll-up are scripts; the interpreter is a generic VBS; **you** run the confirm gate and narrate the failure triage.

---

Shared Resources

| File | Token / call | Purpose | |---|---|---| | `<SAP_DEV_CORE_SHARED_DIR>/rules/safety_policy.md` + `<SAP_DEV_CORE_SHARED_DIR>/scripts/sap_safety_gate.ps1` | Rule 0 | Environment guard — Step 3 runs `-Action assert` before the write | | `<SKILL_DIR>/references/sap_replay_compile.ps1` | `-Action lint\|compile` | Lint (RFC) + segment compiler | | `<SKILL_DIR>/references/sap_replay_exec.vbs` | via 32-bit cscript | Generic segment interpreter (GUI) | | `<SKILL_DIR>/references/sap_replay_table_check.ps1` | `-CheckFile -Values` | RFC_READ_TABLE assertion engine | | `<SKILL_DIR>/references/sap_replay_report.ps1` | `-ReplayFile -CheckFile` | Verdict roll-up + `sapdev.testverdict/1` line | | `<SKILL_DIR>/references/replay_scenario_schema.md` · `example_mm03.replay.json` | docs | Scenario format + example | | `<SAP_DEV_CORE_SHARED_DIR>/rules/test_verdict_contract.md` | contract | The verdict-line shape | | `/sap-gui-probe` · `/sap-gui-skill-scaffold` · `/sap-gui-inspect` | sub-skills | record / scaffold / screenshot-on-fail | | `/sap-login` | sub-skill | Session + broker acquire/release |

---

Step 0 — Resolve Work Directory + Logging

powershell -NoProfile -ExecutionPolicy Bypass -Command ". '<SAP_DEV_CORE_SHARED_DIR>\scripts\sap_settings_lib.ps1'; . '<SAP_DEV_CORE_SHARED_DIR>\scripts\sap_connection_lib.ps1'; Write-Output ('WORK_DIR=' + (Get-SapWorkDir)); Write-Output ('RUN_TEMP=' + (Get-SapRunTemp))"

Then `sap_log_helper.ps1 -Action start -StateFile {RUN_TEMP}\sap_test_replay_run.json`.

`{OUT}` = `Get-SapArtifactDir -ScopeKey TCODE_<tcode> -Skill sap-test-replay` (from `sap_artifact_lib.ps1`; creates the dir) — the artifact output dir the Step 5 roll-up writes to and registers.

Step 1 — Parse Args + init/lint

`run` | `lint` | `init`. `init` (pure local) turns a probe/scaffold folder into a scenario skeleton (steps + guards + popup branches pre-filled, checkpoints stubbed) and stops. `lint`:

C:\Windows\SysWOW64\WindowsPowerShell\v1.0\powershell.exe -ExecutionPolicy Bypass -File "<SKILL_DIR>\references\sap_replay_compile.ps1" -Action lint -Scenario "<scenario>" -Data "<bindings>" -SharedDir "<SAP_DEV_CORE_SHARED_DIR>"

`VERDICT: LINT_ERROR` (unbound token, guard gap, missing tcode/field) -> `REPLAY_SCENARIO_INVALID`, stop. `LINT_PARTIAL` (no RFC) -> proceed with the DDIC checks marked COULD_NOT_CHECK.

Step 2 — run: session + release check

Ensure a GUI session (`/sap-login`), broker `acquire`, liveness check. WARN if the scenario's `recorded_release` != the live `server_release_marker` (guards catch real divergence as REPLAY_ERROR:GUARD).

Step 3 — CONFIRM gate (mandatory; run executes transactions)

**Rule 0 first** (`safety_policy.md`; `run` only — `lint`/`init`/`--dry-run` skip it): `powershell -NoProfile -ExecutionPolicy Bypass -File "<SAP_DEV_CORE_SHARED_DIR>\scripts\sap_safety_gate.ps1" -Action assert -Skill sap-test-replay` — `SAFETY: ALLOW` (0) proceed; `TYPED_CONFIRM_REQUIRED` (3) -> the operator types the shown `PROD <SID>/<CLIENT>` token, re-run with `-ConfirmationText '<their verbatim answer>'`, proceed only on `ALLOW_CONFIRMED`; `REFUSED class=<C>` (1) / `ERROR` (2) -> **STOP**, end `FAILED` with `-ErrorClass <C>`, relay the remediation lines — never bypass or work around it manually. The typed `REPLAY <SID>` / yes-no confirm below still applies after ALLOW/ALLOW_CONFIRMED.

Render tcode, system/client, binding summary, checkpoint count. Non-DEV/modifiable client -> **typed** confirmation `REPLAY <SID>`; DEV -> yes/no. `--dry-run` compiles + lints only (no GUI, no gate).

Step 4 — Compile + segment loop

C:\Windows\SysWOW64\WindowsPowerShell\v1.0\powershell.exe -ExecutionPolicy Bypass -File "<SKILL_DIR>\referen