Skip to main content

What are skills?

Skills are SKILL.md files that you install into your IDE agent. Once installed, your agent understands the full apastra file-based protocol: how to read prompt specs, run evaluations, compare against baselines, and gate quality — all without external infrastructure. Each skill is a focused, composable unit of knowledge. Install only the skills you need, or install everything at once.
Skills work with any agent that supports SKILL.md-based context injection, including Claude Code, Cursor, Amp, Codex, and 37 more IDE agents.

Installation

You can also install individual skills by path:

Available skills

apastra-getting-started

Project setup and onboarding walkthrough. Start here to create your first prompt spec, dataset, evaluator, and suite.

apastra-eval

Run prompt evaluations using your IDE agent as the harness. Supports suite mode and quick eval mode.

apastra-baseline

Establish and manage known-good baselines. Future evals compare against the baseline to detect regressions.

apastra-scaffold

Generate new prompt specs, datasets, evaluators, and suites from templates. All files pass schema validation out of the box.

apastra-validate

Validate all promptops files against JSON schemas. Catch formatting errors and missing fields before running evaluations.

apastra-setup-ci

Upgrade from local-first evaluation to automated GitHub Actions CI with PR gating and governed releases.

Skill reference

Supported IDEs

Skills install into any IDE agent that supports SKILL.md-based context injection. This includes Claude Code, Cursor, Amp, Codex, and 37 more environments. The npx skills CLI handles placement automatically for each IDE.

How skills work

When you install a skill, the npx skills CLI writes the SKILL.md content into your IDE agent’s context directory. From that point on, your agent has the full apastra workflow baked into its knowledge — it knows which files to read, which steps to follow, and which artifacts to produce. The agent becomes your harness. No external runtime, no Python scripts to install, no CI required to get started.
Start with apastra-getting-started to set up your project structure, then add apastra-eval and apastra-baseline for your first eval-and-baseline cycle.