AI agent artifacts
The rule/, skill/, and prompt/ directories contain AI agent artifacts that participate in the standard package inheritance and remix mechanisms. Each directory uses a bundle-based structure where subdirectories group related files for a single artifact.
Understand what belongs in each folder
Rules (rule/)
Rules define behavioral guidelines and constraints for AI agents. Each rule is a directory bundle containing one or more markdown files.
rule/
code-review/
RULE.md
checklist.md
security-audit/
RULE.mdSkills (skill/)
Skills define reusable capabilities for AI agents, including documentation, reference material, and supporting scripts.
skill/
pdf-skill/
SKILL.md
FORMS.md
REFERENCE.md
scripts/
fill_form.py
data-analysis/
SKILL.md
examples/
sample_query.sqlPrompts (prompt/)
Prompts define reusable prompt templates for AI agent interactions.
prompt/
summarize/
PROMPT.md
classify/
PROMPT.md
taxonomy.mdInheritance and remix
These directories follow the same inheritance and remix semantics as all other package categories. A downstream package can:
- Inherit artifacts from dependency packages automatically.
- Override specific files by providing the same path in the downstream package.
- Extend bundles by adding new files to an inherited artifact directory.
Example
# Base package (platform-ai)
platform-ai/
rule/
code-review/
RULE.md
# Application package (my-app depends on platform-ai)
my-app/
rule/
code-review/
RULE.md # Overrides base rules
team-conventions.md # Extends with team-specific content
deploy-review/
RULE.md # New rule not in baseStructural rules
Path conventions
/packageName/rule/bundleName/fileName.ext
/packageName/skill/bundleName/fileName.ext
/packageName/prompt/bundleName/fileName.extFile formats
Any file format is supported within artifact bundles. Common formats include:
.mdfor rules, documentation, and prompt templates..py,.jsfor supporting scripts..json,.yamlfor structured configuration.
Test artifacts
Test-only artifacts live under the test/ overlay:
test/rule/test-review/RULE.md
test/skill/test-skill/SKILL.md
test/prompt/test-prompt/PROMPT.mdRuntime access
Access artifact contents using standard package file APIs:
// List all rule bundles
var files = C3.pkg().files(Pkg.FilesSpec.make({
category: Pkg.Category.RULE
}));
// Read a specific skill file
var content = C3.pkg().content('/my-app/skill/pdf-skill/SKILL.md');
// Check if a path is a rule
var path = Pkg.Path.fromString('meta://my-app/rule/code-review/RULE.md');
path.isRule(); // trueHow AI artifacts relate to other folders
- Source folder (
src/): AI artifacts may reference Types defined insrc/. For example, a skill might document how to use a specific API type. - Resource folder (
resource/): Useresource/for static files that are not AI agent artifacts. Userule/,skill/,prompt/specifically for artifacts intended for AI agent consumption. - Seed folder (
seed/): AI artifacts are not deployed to the database. They remain as package files, unlike seed data which is persisted.