What Artifacts Are Stored in the `outputs/` Directory in AI Engineering From Scratch

The outputs/ directory stores ready-to-use deliverable artifacts—including prompt files, skill bundles, and reference markdown documents—that learners can directly copy into their own projects without running any lesson code.

The rohitg00/ai-engineering-from-scratch repository organizes its curriculum into phases and lessons, with each lesson potentially containing an outputs/ folder that serves as a reference artifact store. According to the source code in site/lesson.html (lines 5274-5279), this directory houses the concrete assets shipped by each lesson, which the site renders in a dedicated "What This Lesson Ships" panel for easy discovery.

Types of Artifacts Found in the outputs/ Directory

The repository defines two primary categories of artifacts within the outputs/ folder. These are validated by the test suite in site/test_build_artifacts.js to ensure consistent structure and usability.

Single-File Prompt Artifacts

These are standalone markdown files containing ready-to-use prompt snippets that illustrate specific techniques taught in the lesson. For example, the skill runtime lesson includes skill-flat-reviewer.md as a concrete prompt artifact. According to site/test_build_artifacts.js (lines 648-656), the test suite explicitly checks for these .md files to ensure they contain valid prompt content that learners can immediately utilize.

Structured Skill Bundles

Complex lessons ship entire directories (bundles) rather than single files. These bundles contain a SKILL.md file plus any supporting resources required to implement the skill within an AI-agent framework. As implemented in site/test_build_artifacts.js (lines 692-696), the validation logic expects bundle directories like release-gate/ to contain a properly formatted SKILL.md file that defines the skill's interface and behavior.

How the Repository Validates outputs/ Content

The repository maintains strict quality controls through automated testing. The file site/test_build_artifacts.js functions as the validation gate for the outputs/ directory structure, ensuring that learners receive functional, well-formed artifacts.

The test suite distinguishes between:

• Markdown artifacts – Individual .md files that must parse correctly as prompt templates
• Bundle artifacts – Subdirectories that must contain a root SKILL.md file and follow the release-gate packaging convention

If the outputs/ folder exists but contains no valid artifacts, the site generation code in site/lesson.html (lines 5274-5279) triggers a fallback UI message indicating that the lesson has no separate output artifacts available.

Accessing and Using outputs/ Artifacts in Your Projects

You can interact with these artifacts programmatically or through simple file operations. The following examples demonstrate how to consume artifacts from the phases/14-agent-engineering/22-skill-runtime/outputs/ path.

Importing a Skill Bundle

import json
import pathlib

# Navigate to the release-gate bundle within the lesson outputs

bundle_path = pathlib.Path(
    "phases/14-agent-engineering/22-skill-runtime/outputs/release-gate"
)

# Load the skill definition (SKILL.md) as referenced in test_build_artifacts.js

skill_definition = (bundle_path / "SKILL.md").read_text()

# Extract the skill description from the first line

print("Loaded skill:", skill_definition.splitlines()[0])

Copying Reference Prompts

# Copy the ready-to-use flat reviewer prompt to your project

cp phases/14-agent-engineering/22-skill-runtime/outputs/skill-flat-reviewer.md \
   my_project/prompts/flat_reviewer.txt

Summary

• The outputs/ directory serves as a reference artifact store for lesson deliverables in the AI Engineering From Scratch curriculum.
• Single-file artifacts (.md files) contain standalone prompts validated by site/test_build_artifacts.js (lines 648-656).
• Bundle artifacts are subdirectories containing SKILL.md and supporting files, verified by tests at lines 692-696 of the same test file.
• site/lesson.html (lines 5274-5279) renders these artifacts in the site's "What This Lesson Ships" UI panel.
• Learners can copy artifacts directly into production projects without executing lesson code.

Frequently Asked Questions

What is the purpose of the outputs/ directory in the AI Engineering From Scratch repository?

The outputs/ directory functions as a curated store of deliverable assets produced by each lesson. According to the repository's source code, these are concrete, ready-to-use files that learners can import into their own AI engineering projects without needing to run the lesson's code or regenerate the assets themselves.

How does the site handle lessons without output artifacts?

When a lesson lacks an outputs/ directory or contains no valid artifacts, the site generation logic in site/lesson.html (lines 5274-5279) automatically displays a fallback message. This informs users that the particular lesson does not ship separate output files, preventing confusion when the "What This Lesson Ships" panel would otherwise be empty.

What file formats are typically found in the outputs/ directory?

The repository stores two primary formats: markdown files (.md) containing prompt templates and documentation, and bundle directories containing a SKILL.md file plus any additional resources required for skill implementation. The test suite site/test_build_artifacts.js validates both formats to ensure they meet the repository's structural requirements.

How can I verify the contents of an outputs/ directory before using it?

You can inspect the outputs/ folder directly in the file system or consult the validation logic in site/test_build_artifacts.js. This test file defines the expected structure—either a standalone markdown file like skill-flat-reviewer.md or a bundle directory like release-gate/ containing SKILL.md—providing a reliable reference for what constitutes a valid artifact.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}