Aerospike agent skills

This repository holds Aerospike-related Agent Skills under skills/. Each skill is a self-contained folder you can copy into Cursor, add to GitHub Copilot context via repo instructions, or use with Claude and other assistants. Skills focus on the core Aerospike Database (not Aerospike Graph).

Plain-language topics: real-time / low-latency key-value storage, Redis or Memcached–style patterns with persistence, feature stores, user profiles, client-side modeling and policies.

The canonical list of skills (and what to copy) is skills/README.md. More Aerospike skills may appear there over time.

Layout note: If you previously used aerospike-getting-started/ at the repository root, it now lives at skills/aerospike-getting-started/. That folder is for contributors who edit the skill sources; users install the compiled skill from compiled-skills/.

Skills in this repository

Choosing a skill: Use getting-started when you need a dev instance, correct defaults, or a first verified read/write. Use development when you are writing or reviewing client code against a model that already exists. Use data-modeling when the schema itself is the deliverable. You can install all; they complement each other.

Out of scope for these skills: production cluster topology, full operations runbooks, Kubernetes/Helm, or cloud-specific deployment—for that, use Aerospike documentation (links also appear in the getting-started skill).

Who this is for

• Developers who want a working local instance and first client I/O (getting-started).
• Developers building or reviewing Aerospike client applications (development skill).
• Teams that want consistent AI-assisted answers when working with Aerospike.

What’s in the repo

Skill packaging evaluation (structure A/B tests, coverage gates) runs in a separate harness repository and is not part of this repository's checks.

The canonical aerospike.conf and commands for local setup live in skills/aerospike-getting-started/SKILL.md (not duplicated here).

Using with AI assistants

We publish one skill: compiled-skills/aerospike/SKILL.md, compiled from the three authoring folders under skills/. Registries list that artifact only. The folders under skills/ are maintained for editing and compilation, not as separate install targets.

The skill and its rule files ship together — install the directory, not the single file, or the rules SKILL.md names will be missing. It is semantically versioned: each release is tagged vX.Y.Z, and an installed copy states its own version in metadata.version so you can tell what you have without checking back here.

Use without AI tools (any editor)

Path A: First local instance and first put/get

1. Read Critical rules and Hallucination blacklist in skills/aerospike-getting-started/SKILL.md.
2. Run Autonomous first-run steps in the same file (Docker + config + verify logs).
3. Pick your language in examples.md and run a write/read test.
4. For Compose, Enterprise evaluation, or troubleshooting, see reference.md.

Path B: Application development (modeling, CDTs, policies)

1. Read skills/aerospike-development/SKILL.md for scope and mental model.
2. Use skills/aerospike-development/references/README.md for the rules and example TOC; reference.md and examples.md as needed.

First-hour checklist (getting-started skill)

Use this when your goal is a running local database and a verified first client write/read.

• Docker is installed and docker --version works.
• Aerospike container is running per skills/aerospike-getting-started/SKILL.md (ports 3000–3002 mapped; add 3003 if you need admin access on Database 8.1.0+—see that file).
• Logs show the ready line described in SKILL.md (successful startup).
• Client SDK installed for your language; write/read example prints success.
• (Optional) Skim Documentation links in the same SKILL.md for next steps.

Maintenance and distribution

• metadata.last_verified: Each skill’s SKILL.md may carry last_verified under metadata. When you change Docker images, client commands, or major facts for that skill, re-check the documented flow and update the date.
• Linting skills: Run ./scripts/validate-skill.sh (see CONTRIBUTING.md); CI runs the same check via .github/workflows/skill-validator.yml.
• Spec conformance: Run ./scripts/validate-spec.sh to check SKILL.md frontmatter against the Agent Skills specification; CI runs it via .github/workflows/spec-conformance.yml.
• Publishing: Registry submission is automated on release—see docs/PUBLISHING.md.
• Compiled skill: compiled-skills/ is generated from skills/. After editing any skill, run python3 scripts/compile-agents.py --write; CI fails the PR if the compiled output is stale. Frontmatter for the published skill lives in scripts/skills_compile/published_skill.yaml.
• Contributing: See CONTRIBUTING.md for how to add or edit skills and keep root docs in sync.
• Redistribution: This repository is licensed under the Apache License 2.0, matching aerospike/data-modeling-guide.

Official documentation

Links to Aerospike docs (quick start, configuration, clients, Docker) are collected in the Documentation links section at the bottom of skills/aerospike-getting-started/SKILL.md. The development skill’s reference.md points to additional client and operations documentation where relevant.