sap-system-personality-report

name	sap-system-personality-report
description	Generate a comprehensive SAP System Personality Report. Analyzes custom code landscape, functional footprint, development activity, health metrics, and package breakdown. Use when a user asks to characterize a system, understand a system, get a system overview, system report, system personality, or "what does this system do?" Collects data via SQL queries and presents results in a structured webview.
argument-hint	[connectionId of the SAP system to analyze]
user-invocable	true
disable-model-invocation	false

SAP System Personality Report

Generate a rich, structured report that characterizes an SAP system in human-readable terms. You will collect data by running SQL queries, then present the results in a webview table AND provide a narrative AI summary.

CRITICAL: Before running ANY query, call get_abap_sql_syntax to understand ABAP SQL syntax. Use get_object_lines to check table structures before running SQL queries.

CRITICAL: All queries must use ABAP SQL syntax (e.g., ORDER BY field DESCENDING not DESC). Use the execute_data_query tool with displayMode: "internal" to collect data, then assemble final results into a displayMode: "ui" call.

---

Step 1: System Identity

Get basic system information first. Use the get_sap_system_info tool — it returns system type (S/4HANA vs ECC), release, timezone, and client details. No SQL needed for this step.

---

Step 2: Custom Code Landscape — Object Counts

Run these queries to count custom objects. Use displayMode: "internal" with rowRange: {start: 0, end: 5} since we only need counts.

Run all of these in parallel (they are independent):

Classes

SELECT COUNT(*) AS CNT FROM SEOCLASS WHERE CLSNAME LIKE 'Z%' OR CLSNAME LIKE 'Y%'

Interfaces

NOTE: The SEOINTERF table does NOT exist in S/4HANA. Use TADIR instead:

SELECT COUNT(*) AS CNT FROM TADIR WHERE PGMID = 'R3TR' AND OBJECT = 'INTF' AND OBJ_NAME LIKE 'Z%'

Then separately for Y:

SELECT COUNT(*) AS CNT FROM TADIR WHERE PGMID = 'R3TR' AND OBJECT = 'INTF' AND OBJ_NAME LIKE 'Y%'

Sum both counts.

Programs/Reports (by type)

CRITICAL: The TRDIR table contains ALL program types (reports, includes, class pools, etc.). Using a raw count is very misleading since includes (SUBC='I') dominate the count. Always break down by SUBC field:

SELECT SUBC, COUNT(*) AS CNT FROM TRDIR WHERE NAME LIKE 'Z%' OR NAME LIKE 'Y%' GROUP BY SUBC ORDER BY CNT DESCENDING

SUBC values: '1'=Executable Reports, 'I'=Includes, 'M'=Module Pools, 'S'=Subroutine Pools, 'K'=Class Pools, 'J'=Interface Pools, 'X'=XSLT Programs.

In the final report, present these as separate line items:

• Executable Reports (SUBC='1') — the actual standalone reports
• Module Pools (SUBC='M') — dialog programs
• Includes (SUBC='I') — sub-objects (generated for classes, FMs, etc.) — show for context but note they are generated
• Subroutine Pools (SUBC='S') — standalone subroutine containers

Omit class pools (K) and interface pools (J) from the report since those are already counted under Classes and Interfaces.

Function Modules

SELECT COUNT(*) AS CNT FROM TFDIR WHERE FUNCNAME LIKE 'Z%' OR FUNCNAME LIKE 'Y%'

Function Groups

SELECT COUNT(*) AS CNT FROM TLIBG WHERE AREA LIKE 'Z%' OR AREA LIKE 'Y%'

Database Tables (transparent tables only)

SELECT COUNT(*) AS CNT FROM DD02L WHERE TABNAME LIKE 'Z%' AND TABCLASS = 'TRANSP' AND AS4LOCAL = 'A'

Then separately for Y:

SELECT COUNT(*) AS CNT FROM DD02L WHERE TABNAME LIKE 'Y%' AND TABCLASS = 'TRANSP' AND AS4LOCAL = 'A'

Sum both counts.

Structures

SELECT COUNT(*) AS CNT FROM DD02L WHERE TABNAME LIKE 'Z%' AND TABCLASS = 'INTTAB' AND AS4LOCAL = 'A'

Then separately for Y and sum.

Data Elements

SELECT COUNT(*) AS CNT FROM DD04L WHERE ROLLNAME LIKE 'Z%' AND AS4LOCAL = 'A'

Then separately for Y and sum.

Domains

SELECT COUNT(*) AS CNT FROM DD01L WHERE DOMNAME LIKE 'Z%' AND AS4LOCAL = 'A'

Then separately for Y and sum.

Table Types

SELECT COUNT(*) AS CNT FROM DD40L WHERE TYPENAME LIKE 'Z%' AND AS4LOCAL = 'A'

Then separately for Y and sum.

CDS Views

SELECT COUNT(*) AS CNT FROM DDDDLSRC WHERE DDLNAME LIKE 'Z%'

Then separately for Y and sum.

Message Classes

SELECT COUNT(*) AS CNT FROM T100A WHERE ARBGB LIKE 'Z%' OR ARBGB LIKE 'Y%'

Custom Transactions

SELECT COUNT(*) AS CNT FROM TSTC WHERE TCODE LIKE 'Z%' OR TCODE LIKE 'Y%'

Enhancement Implementations (BAdIs)

Shows how heavily the system extends SAP standard:

SELECT COUNT(*) AS CNT FROM TADIR WHERE PGMID = 'R3TR' AND OBJECT = 'ENHO' AND (OBJ_NAME LIKE 'Z%' OR OBJ_NAME LIKE 'Y%')

Number Range Objects

SELECT COUNT(*) AS CNT FROM TADIR WHERE PGMID = 'R3TR' AND OBJECT = 'NROB' AND (OBJ_NAME LIKE 'Z%' OR OBJ_NAME LIKE 'Y%')

NOTE: Some queries may fail on specific systems (missing tables, authorization). If a query fails, record the count as "N/A" and continue with the rest.

---

Step 3: Package Breakdown

This is the most valuable section. Get object counts per custom package.

CRITICAL: The column OBJECT is a reserved keyword in ABAP SQL. Do NOT alias it (e.g., OBJECT AS OBJ_TYPE will cause a parser error Unknown column name "O"). Use the column name OBJECT directly without alias.

SELECT DEVCLASS, OBJECT, COUNT(*) AS OBJ_COUNT
FROM TADIR
WHERE DEVCLASS LIKE 'Z%'
  AND PGMID = 'R3TR'
GROUP BY DEVCLASS, OBJECT
ORDER BY DEVCLASS ASCENDING

Use displayMode: "internal" with a large enough rowRange (start: 0, end: 1000) and maxRows: 5000.

Note: We no longer filter by specific OBJECT types in the WHERE clause — instead retrieve all R3TR objects and filter/pivot in post-processing. This avoids issues with long IN() lists and captures all object types.

Then run the same for Y packages:

SELECT DEVCLASS, OBJECT, COUNT(*) AS OBJ_COUNT
FROM TADIR
WHERE DEVCLASS LIKE 'Y%'
  AND PGMID = 'R3TR'
GROUP BY DEVCLASS, OBJECT
ORDER BY DEVCLASS ASCENDING

Pivot the results into a table with columns: | Package | Classes | FMs | Reports | DD Objects | CDS | Interfaces | Other | Total |

Where "DD Objects" = sum of TABL + DTEL + DOMA + TTYP + VIEW + ENQU + SHLP for that package.

Sort by Total descending. Show top 30 packages.

---

Step 4: Development Timeline & Object Quality

Oldest and newest custom objects

Shows when custom development started and when the most recent object was created:

SELECT MIN( CREATED_ON ) AS OLDEST, MAX( CREATED_ON ) AS NEWEST
FROM TADIR
WHERE (OBJ_NAME LIKE 'Z%' OR OBJ_NAME LIKE 'Y%')
  AND PGMID = 'R3TR'
  AND CREATED_ON <> '00000000'

Inactive objects

Objects changed but never activated — a code quality smell:

SELECT COUNT(*) AS CNT FROM DD02L WHERE (TABNAME LIKE 'Z%' OR TABNAME LIKE 'Y%') AND AS4LOCAL = 'M'

SELECT COUNT(*) AS CNT FROM DD04L WHERE (ROLLNAME LIKE 'Z%' OR ROLLNAME LIKE 'Y%') AND AS4LOCAL = 'M'

AS4LOCAL = 'M' means modified/inactive. Sum the counts across DD02L and DD04L for a rough inactive object count. If the count is high relative to active objects, it's a quality concern worth noting.

---

Step 5: Development Activity (Last 90 Days)

Active developers and transport counts

SELECT AS4USER, COUNT(*) AS TR_COUNT
FROM E070
WHERE AS4DATE >= '<CALCULATE: today minus 90 days, format YYYYMMDD>'
  AND STRKORR <> ''
GROUP BY AS4USER
ORDER BY TR_COUNT DESCENDING

Date calculation: You must compute the date yourself. Today's date minus 90 days, formatted as YYYYMMDD (e.g., 20251205 for Dec 5, 2025). ABAP SQL dates are stored as YYYYMMDD strings.

Use maxRows: 100, rowRange: {start: 0, end: 50}.

Total transport count

SELECT COUNT(*) AS TOTAL
FROM E070
WHERE AS4DATE >= '<90 days ago YYYYMMDD>'
  AND STRKORR <> ''

---

Step 6: System Health (Last 30 Days)

Dump trend

SELECT DATUM, COUNT(*) AS DUMP_COUNT
FROM SNAP
WHERE DATUM >= '<CALCULATE: today minus 30 days, format YYYYMMDD>'
GROUP BY DATUM
ORDER BY DATUM ASCENDING

Top dump types

CRITICAL: The SNAP table in S/4HANA does NOT have a FESSION (or SEESSION) field for error type. The S/4HANA SNAP table only contains: DATUM, UZEIT, AHOST, UNAME, MANDT, MODNO, SEQNO, XHOLD, and FLIST01-08 (text blobs).

Use the SNAP_ADT table instead — it has a structured RUNTIME_ERROR field with the error type name:

SELECT RUNTIME_ERROR, COUNT(*) AS CNT
FROM SNAP_ADT
WHERE DATUM >= '<30 days ago YYYYMMDD>'
GROUP BY RUNTIME_ERROR
ORDER BY CNT DESCENDING

Use maxRows: 20, rowRange: {start: 0, end: 10}.

Test class count (rough indicator)

SELECT COUNT(*) AS CNT FROM SEOCLASS WHERE CLSNAME LIKE 'Z%TEST%' OR CLSNAME LIKE 'Y%TEST%' OR CLSNAME LIKE 'ZCL%TEST%' OR CLSNAME LIKE 'ZCL_TEST%'

NOTE: SNAP table may not be accessible due to authorization. On S/4HANA, the SNAP table has a simplified structure — use SNAP_ADT for error type breakdowns instead. If both SNAP and SNAP_ADT queries fail, skip the health section and note "Dump data unavailable — likely authorization restriction."

---

Step 7: Assemble and Present

Determine Functional Footprint

From the package breakdown data collected in Step 3, categorize packages by SAP functional area.

Use your own knowledge to categorize: Look at package names, their descriptions (if available), and the types of objects they contain. Common patterns:

• SD: sales, order, delivery, billing, pricing, shipping
• MM: material, purchasing, procurement, inventory, vendor
• FI: finance, accounting, payment, bank, tax, invoice
• CO: controlling, cost, profit, overhead
• PP: production, manufacturing, plant, BOM, routing
• QM: quality, inspection, lot
• PM: maintenance, equipment, notification
• WM/EWM: warehouse, storage, bin
• HR/HCM: personnel, payroll, time, absence
• BC/BASIS: tools, utilities, framework, logging, middleware

Group packages into these areas and calculate percentages.

Build the Final Display

Use execute_data_query with displayMode: "ui" and data parameter (direct data input) to create the final report as a structured table.

Create ONE comprehensive display with these sections as a data table:

Title: "System Personality Report: {connectionId}"

Build a summary data table with columns:

Category	Metric	Value

Rows should include:

• System type, release, timezone
• Development timeline: oldest custom object date → newest custom object date
• Each object type count broken down properly:
  • Executable Reports (SUBC='1'), Module Pools (SUBC='M'), Includes (SUBC='I' — note as "generated/sub-objects"), Subroutine Pools (SUBC='S')
  • Classes, Interfaces, Function Modules, Function Groups
  • DB Tables, Structures, Data Elements, Domains, Table Types, CDS Views, Message Classes
  • Custom Transactions, Enhancement Implementations, Number Range Objects
• Total custom objects (exclude includes and class/interface pools from the total to avoid double-counting)
• Inactive objects count (if > 0, note as quality concern)
• Test class count and coverage %
• Top 5 packages by object count
• Active developers count and top 5 by transport count
• Total transports (90 days)
• Average dumps per day (30 days)
• Top 3 dump types

AI Narrative Summary

After collecting all data, write a 5-20 sentence narrative paragraph summarizing the system's personality. Include:

1. What kind of company/industry this appears to be (inferred from functional footprint)
2. Scale of custom development and how long it's been going (development timeline)
3. Team size and activity level
4. Code quality indicators (dump rate trend, test coverage, inactive object count)
5. Enhancement footprint — how heavily standard SAP is extended
6. Any notable patterns (heavy CDS adoption = modernization, few tests = risk, one developer doing 50% of work = key-person dependency, many inactive objects = cleanup needed, many custom transactions = heavy user-facing customization)
7. One actionable recommendation

Present this narrative to the user as part of your response text, alongside the data table.

---

Error Handling

• If a query fails, skip that section and note it in the report. Never let one failed query block the entire report.
• If the system has zero custom objects, say so: "This system has no custom Z/Y development."
• If SNAP is inaccessible, skip health. If E070 is inaccessible, skip activity.
• Always present whatever data you successfully collected.

---

Performance Notes

• All queries are read-only SELECTs on metadata tables — very lightweight
• Run independent queries in parallel when possible (Steps 2 queries can all run simultaneously)
• Total data collection should take 10-30 seconds depending on system
• Package breakdown (Step 3) is the heaviest query but still fast since it's aggregated