How Oracle Skills Open Source Signals the Rise of the AI Skill Era

From Personal Project to Official Project

In early March 2026, Kris Rice, senior vice president of product development at Oracle, released 117 independent reference guides in his personal repository, covering SQL/PLSQL development, performance tuning, security, management, monitoring, architecture, DevOps, migration and more across 14 categories. Each guide includes actionable examples, best practices, common pitfalls and explicit version‑compatibility notes as defined in SKILL_AUTHORING_GUIDE.md.

The project was later merged into the official Oracle organization, gaining an Oracle maintenance commitment. Jeff Smith stated, “we plan to maintain these, add more, and they should be good for all supported versions of Oracle AI Database (19c, 23ai, 26ai).” This assures enterprise users that the skills will remain maintained.

Commit history shows multiple validations against official Oracle documentation to mitigate large‑model hallucinations.

Oracle Skills as a New "Interface"

When first opening the Oracle Skills repository, you will not find complex code, business logic, or an SDK. Instead you see a collection of SKILL.md files that define how an AI should perform a capability, including input, output, behavior constraints and usage.

[shawnyan@rl9 oracle-skills]$ cat ./db/SKILL.md
---
name: db
description: Oracle Database skills for administration, SQL and PL/SQL development, performance tuning, security, ORDS, SQLcl, migrations, frameworks, Oracle Container Registry guidance, and agent‑safe database workflows.
---

# Oracle Database Skills

This domain contains Oracle Database skills for administration, SQL and PL/SQL development, performance tuning, security, ORDS, SQLcl, migrations, frameworks, OCR container guidance, and agent‑safe database workflows.

## How to Use This Domain

1. Start with the routing table below.
2. Read only the specific file or category you need.

## Directory Structure
...
## Category Routing
...
## Key Starting Points
...
## Common Multi‑Step Flows
...

In traditional engineering, an API is the interface, an SDK the toolkit, a function call the trigger, and parameters the input. In the AI world, a Skill is the interface, a Prompt template the toolkit, a Skill Invocation the trigger, and a Context (files + instructions) the input. Prompts become structured, reusable, engineering‑grade capability interfaces.

Multi‑Domain Architecture

Oracle Skills adopts a clear multi‑domain design, each domain mapping to a technology area:

oracle/skills/
├── db/          ← core database domain, 100+ skills
│   ├── SKILL.md
│   ├── admin/          # backup/restore, RMAN, Data Guard, user management
│   ├── agent/          # AI Agent safety workflows
│   ├── appdev/         # JDBC, connection pools, JSON/XML, transaction management
│   ├── architecture/   # RAC, Multitenant, Exadata, In‑Memory, OCI
│   ├── containers/    # Oracle Container Registry guidance
│   ├── design/         # ERD, data modeling, partitioning, tablespaces
│   ├── devops/         # Liquibase, CI/CD, online operations, version control
│   ├── features/      # AQ, Scheduler, materialized views, DBLinks, APEX
│   ├── frameworks/    # SQLAlchemy, Django, Spring JPA, TypeORM, etc.
│   ├── migrations/     # PG/MySQL/SQL Server → Oracle migration
│   ├── monitoring/     # Alert Log, ADR, space, Top SQL
│   ├── ords/           # Oracle REST Data Services
│   ├── performance/    # AWR, ASH, execution plans, indexes, optimizer
│   ├── plsql/          # packages, cursors, collections, error handling, unit tests
│   ├── security/       # privileges, VPD, TDE, auditing, network security
│   ├── sql-dev/        # SQL tuning, window functions, CTE, dynamic SQL
│   └── sqlcl/          # SQLcl CLI, Liquibase, MCP Server
├── oci/          # Oracle Cloud Infrastructure (reserved)
├── fusion/       # Oracle Fusion (reserved)
├── apex/         # Oracle APEX (reserved)
└── graal/        # GraalVM Native Image (in progress)

The db/agent/ sub‑directory exemplifies Oracle’s “safe‑driving manual” for AI Agents, containing skills such as: safe-dml-patterns.md – enforces Always‑WHERE guards, COUNT‑first checks, SAVEPOINT dry‑runs before UPDATE/DELETE. destructive‑op‑guards.md – pre‑check lists and impact analysis for DROP, TRUNCATE, etc., preventing accidental data loss. idempotency‑patterns.md – ensures DDL/DML operations are idempotent for reliable CI/CD deployments. nl-to-sql-patterns.md and intent-disambiguation.md – define natural‑language‑to‑SQL translation strategies, e.g., interpreting “increase sales department salaries by 10%” into a safe UPDATE.

Jeff Smith’s test case shows the safety mechanism in action: when the agent was asked to “give employees a 10% raise or $5,000, whichever is lower,” the safe‑dml pattern forced a COUNT query first, then applied the UPDATE only after the WHERE clause was validated.

Installation

Oracle Skills can be installed via the standard npx skills CLI, compatible with Claude Code, Cursor, Cline and other AI coding assistants.

# Install all skills
npx skills add oracle/skills

# Install only the database domain
npx skills add oracle/skills/db

Chinese users may download a ZIP package from https://gitcode.com/mydb/oracle-skills.

Skill Document Structure

Each skill document follows a uniform template: Overview, Practical Examples (copy‑able code), Best Practices (official recommendations), Common Mistakes (pitfalls), Oracle Version Notes (compatibility), and Sources (citations). This self‑contained design lets an AI Agent load only the needed skill files, saving tokens and avoiding context pollution.

Version coverage is pragmatic: 19c is the baseline compatibility target; new‑version features must be explicitly marked with a 19c‑compatible fallback. This serves both long‑term‑support customers and early‑adopter innovators.

Why the "Skill Era" Is Arriving

The repository’s value lies not in code but in defining a new paradigm. Over the past decade software engineering has moved from functions to services to APIs; today the emerging abstraction is the Skill. Skills encapsulate AI‑driven capabilities as reusable, composable modules, enabling agents to invoke stronger models, browsers or external systems before reaching the final model.

In this model, an Agent calls a Skill, which may call another model or tool, forming an Agent → Skill → Model chain. Engineers shift from writing functions and handling API responses to orchestrating Skills and composing capabilities. Skills act like npm, Maven or PyPI packages, but they deliver abilities rather than code.

From a database perspective, the shift is profound: traditionally humans write SQL that the database executes. With Skills, humans and Agents invoke a Skill that generates SQL automatically, turning SQL into an intermediate artifact rather than a direct entry point. Future databases may be understood and participate in reasoning, affecting query optimization, schema design and data analysis.

Practical Guidance

For DBAs, prioritize installing admin, monitoring, performance and security skills to let AI assist with inspections and diagnostics, but enable read‑only suggestion mode for agent‑executed permissions. For AI application developers, the ai-profiles, vector-search and dbms-vector skills are essential; 26ai’s AI Vector Search with SELECT AI forms a killer combination, and the repository includes a complete RAG construction guide. DevOps or platform engineers can use devops, sqlcl, containers and agent/idempotency‑patterns to bring database changes into a GitOps workflow, achieving true Database‑as‑Code.

Conclusion

As an experienced DBA, the author acknowledges that while AI Agents can retrieve hundreds of Oracle‑validated expert guides in seconds, the traditional manual approach feels inefficient. Oracle Skills does not replace DBAs; it amplifies their capability radius. Adopt it wisely, but retain human oversight, especially for safety‑critical operations.

References

Oracle Skills official repository: https://github.com/oracle/skills

Jeff Smith’s test blog: https://www.thatjeffsmith.com/archive/2026/03/having-a-go-with-100-new-ai-skills-for-oracle-ai-database/

Oracle Database / Oracle AI Database 26ai documentation: https://docs.oracle.com/en/database/oracle/oracle-database/26/index.html