A collection of AI agent skills for SAP ABAP engineering — built by a SAP consultant for daily real-world work.
sap-engineering-skill is a monorepo of AI agent skills that cover the core workflow of SAP ABAP development: reading and writing source code, reviewing code quality and security before release, assessing transport requests at the gate, and answering SAP integration questions with production-ready precision.
Each skill follows the standard SKILL.md specification and works with any compatible AI agent framework — opencode, Claude Code, Cursor, or any framework that supports custom tool/skill injection.
The three subtree skills are also maintained as independent public repositories so they can be used standalone.
A command-line tool and AI agent skill for reading and writing ABAP source code, metadata, and transport requests from SAP systems via the ADT REST API.
Supports: programs, classes, function modules, function groups, interfaces, includes, CDS views, DDIC tables/structures/types, packages, transactions, SQL queries, where-used analysis, syntax checks, transport management.
Key features:
- Read-only by default; write and transport operations require explicit capability flags + per-operation
[y/N]confirmation - Windows one-click installer (
setup-opencode-abap-cli.bat) that wires up opencode end-to-end - Credentials stored at
~/.sap-adt-cli/config.json; environment variable override for CI/CD
An AI agent skill for SAP ABAP pre-release code review. Performs a comprehensive security and quality assessment across 9 dimensions and produces a formal, sign-off-ready Markdown report.
| # | Dimension | Focus |
|---|---|---|
| 1 | [SEC] Security | SQL injection, code injection, hardcoded credentials |
| 2 | [AUTH] Authorization | Missing AUTHORITY-CHECK, bypass patterns |
| 3 | [DATA] Data Integrity | SY-SUBRC handling, locking, exception handling |
| 4 | [PERF] Performance | SELECT-in-LOOP, SELECT *, full table scans |
| 5 | [STD] Code Standards | Deprecated statements, oversized methods, dead code |
| 6 | [INTERFACE] Integration | Dialog in RFC FMs, missing EXCEPTIONS, OData auth |
| 7 | [CHANGE] Change Impact | Affected tables, SAP standard modifications |
| 8 | [COMP] Compliance | PII, audit logs, SoD paths |
| 9 | [FUNC] Functional (optional) | Business scenario coverage vs. requirements |
Output: GO / CONDITIONAL GO / NO-GO recommendation with evidence-cited findings and a sign-off table.
An AI agent skill that performs structured, evidence-driven release readiness assessment for SAP Transport Requests. Produces an auditable GO / CONDITIONAL_GO / NO_GO / NEED_MORE_EVIDENCE decision.
Covers 10 review dimensions: code quality, performance, security, authorization, transaction consistency, integration impact, transport completeness, functional alignment, release readiness, and evidence gaps.
Three review modes:
- Offline Package — structured Review Package exported from SAP (preferred)
- Offline Local — partial materials (source files only)
- Online Transport — TR ID +
tr_collector.pyCLI for live ADT collection
Core principle: Evidence-first. AI never invents conclusions from insufficient evidence.
A composable knowledge-base skill that turns any AI assistant into a SAP integration specialist. Covers 9 business domains and 8 integration technologies — no more generic wrong answers.
Business domains: MM (Purchasing, Inventory), SD (Sales), FI (GL, AR/AP, Asset Accounting, FSSC), Master Data, PP (Production)
Technologies: OData V2/V4, RFC/JCo, SOAP over HTTP RFC, IDoc/PI-PO, BAPI & RAP, Authentication, BTP Integration Suite, Best Practices
SAP versions: ECC 6.0 · S/4HANA On-Prem 1909–2023+ · S/4HANA Cloud (Public & Private Edition)
sap-engineering-skill/
├── README.md ← This file (English)
├── README.zh-CN.md ← Chinese version
├── LICENSE
├── setup-opencode-abap-cli.bat ← Windows one-click installer
└── skills/
├── sap-adt-cli/ ← ADT CLI tool & skill (source in this repo)
├── abap-code-review/ ← ABAP code review skill [git subtree]
├── sap-transport-gate/ ← Transport gate review skill [git subtree]
└── sap-integration-wiki/ ← SAP integration knowledge base [git subtree]
The three subtree skills are tracked via named git remotes:
| Remote | Repository |
|---|---|
pub-abap-code-review |
https://github.com/shrek-abaper/abap-code-review |
pub-sap-transport-gate |
https://github.com/shrek-abaper/sap-transport-gate |
pub-sap-integration-wiki |
https://github.com/shrek-abaper/sap-integration-wiki |
Download setup-opencode-abap-cli.bat and double-click. The script installs opencode, clones this repo, and wires up the skill automatically.
# Clone the repository
git clone https://github.com/shrek-abaper/sap-engineering-skill
cd sap-engineering-skill
# Link skills into your agent's skill directory
ln -s "$(pwd)/skills/sap-adt-cli" ~/.agents/skills/sap-adt-cli
ln -s "$(pwd)/skills/abap-code-review" ~/.agents/skills/abap-code-review
ln -s "$(pwd)/skills/sap-transport-gate" ~/.agents/skills/sap-transport-gate
ln -s "$(pwd)/skills/sap-integration-wiki" ~/.agents/skills/sap-integration-wiki
# Configure SAP credentials for sap-adt-cli
python3 skills/sap-adt-cli/scripts/sap_adt_cli.py configure| Agent | Notes |
|---|---|
| opencode | Free, open-source. Supports 30+ model providers. Windows installer included. |
| Claude Code | Anthropic's official CLI agent — natively supports SKILL.md |
| Cursor | AI-powered code editor; install via MCP tool adapter |
⚠️ Data compliance: ABAP source code may contain core business logic and sensitive data. Before sending code to cloud-based AI services, confirm compliance with your organization's data security policy.
| Skill | Use When |
|---|---|
sap-adt-cli |
Read/write ABAP source, run SQL, manage transports via ADT API |
abap-code-review |
Pre-release security & quality review of a single ABAP program (9 dimensions) |
sap-transport-gate |
TR-level release gate assessment — evidence-based GO/NO-GO decision |
sap-integration-wiki |
SAP integration patterns, API reference, troubleshooting by scenario |
This project was built using an AI-native, spec-driven development loop:
- Specification — Requirements were worked out in conversation with Claude Sonnet 4.6, producing structured Markdown prompt files encoding the full spec (scope, acceptance criteria, anti-patterns).
- Execution — Prompt files were consumed end-to-end by opencode + Oh My OpenAgent running on a GitHub Copilot Pro subscription — which planned, implemented, tested, and verified changes autonomously.
- Review — The human role was limited to deciding what to build, reviewing outputs, and iterating on the spec.
The author is a SAP consultant with over ten years of project experience and no software development background. This project grew from a practical daily need and became a personal experiment in how an AI-native workflow can reshape what domain experts are able to build — shared here in the hope it's useful to others on a similar path.
Skills contain reference content from:
- SAP Clean ABAP Style Guide — CC BY 4.0 (used in
abap-code-review) - SAP Business Accelerator Hub, SAP Help Portal — public SAP documentation (used in
sap-integration-wiki)
This project is not affiliated with, endorsed by, or officially supported by SAP SE.