CI/CD Integration

Relevant source files

The following files were used as context for generating this wiki page:

• packages/core/src/cicd/index.ts
• packages/core/src/cicd/templates.ts
• e2e/commands-settings.e2e.test.ts
• packages/core/src/methodology/packs/planning/verification-gates/SKILL.md
• README.md

The CI/CD Integration module in SkillKit provides a robust framework for automating the validation, testing, and deployment of AI agent skills. Its primary purpose is to ensure that specialized instruction files (SKILL.md) maintain high quality and structural integrity as they are modified or added to a repository. By integrating with popular DevOps platforms like GitHub, GitLab, and CircleCI, the system allows teams to enforce "Verification Gates" that prevent faulty or non-compliant skills from reaching production agents.

This module works in tandem with the Skill Testing Framework and Verification Gates to provide automated checkpoints. It leverages the skillkit validate and skillkit test commands to perform static analysis and functional verification within automated pipelines.

Sources: packages/core/src/cicd/index.ts:1-7, README.md:144-150, packages/core/src/methodology/packs/planning/verification-gates/SKILL.md:1-15

System Architecture and Workflow

The CI/CD integration follows a template-driven architecture where pre-defined configurations are injected into the local development environment or remote pipeline runners. The workflow typically begins with local validation via pre-commit hooks and culminates in comprehensive testing within a remote CI environment.

CI/CD Workflow Diagram

The following diagram illustrates the flow from a developer's local commit to the final validation in a remote repository.

Automation Templates

SkillKit provides built-in templates for various platforms. These templates are accessible via the skillkit cicd command and are designed to minimize the configuration overhead for developers.

Supported Platforms and Templates

Platform	Template Variable	File Type	Purpose
GitHub Actions	GITHUB_ACTION_TEMPLATE	YAML	Automates validation on push/PR to specific skill paths.
GitLab CI	GITLAB_CI_TEMPLATE	YAML	Configures test stages and artifact collection for GitLab.
CircleCI	CIRCLECI_CONFIG_TEMPLATE	YAML	Defines Docker-based jobs for skill validation.
Git Hooks	PRE_COMMIT_HOOK_TEMPLATE	Bash	Runs local checks before code is committed to Git.
pre-commit	PRE_COMMIT_CONFIG_TEMPLATE	YAML	Integration for the pre-commit.com framework.

Sources: packages/core/src/cicd/templates.ts:11-235, e2e/commands-settings.e2e.test.ts:303-345

Template Retrieval Logic

The getCICDTemplate function serves as the central API for retrieving these configurations based on a provider name.

export function getCICDTemplate(name: string): string | null {
  switch (name.toLowerCase()) {
    case 'github':
    case 'github-action':
      return GITHUB_ACTION_TEMPLATE;
    case 'gitlab':
    case 'gitlab-ci':
      return GITLAB_CI_TEMPLATE;
    // ... handles other cases
  }
}

Sources: packages/core/src/cicd/templates.ts:241-262

Implementation of Verification Gates

Verification gates are formalized checkpoints within the CI/CD pipeline. These gates categorize validation into automated and manual processes to ensure multifaceted quality control.

Automated CI Gates

The following logic is embedded within the provided CI templates to enforce quality:

1. Static Validation: Executes skillkit validate to ensure YAML frontmatter is present, names are lowercase, and descriptions meet length requirements.
2. Functional Testing: Executes skillkit test --json to run assertions defined for the skills.
3. Result Analysis: Uses tools like jq to parse JSON test outputs and determine the exit status of the pipeline.

Sources: packages/core/src/cicd/templates.ts:46-70, packages/core/src/methodology/packs/planning/verification-gates/SKILL.md:105-125

Local Pre-commit Hook Flow

The local pre-commit hook acts as the first line of defense. It identifies staged files and only runs validation if changes are detected in relevant directories (e.g., .claude/skills/, .cursor/skills/).

Configuration and Usage

Users initialize CI/CD integration through the CLI. The command skillkit cicd init --provider <name> generates the necessary configuration files in the project root or relevant subdirectories (e.g., .github/workflows/).

CLI Command Options

Option	Type	Description
--provider	string	The CI/CD platform (e.g., github, gitlab, circleci).
--dry-run	boolean	Previews the template without writing to disk.
--json	boolean	Outputs template information in JSON format.

Sources: e2e/commands-settings.e2e.test.ts:315-345, packages/core/src/cicd/templates.ts:267-275

Conclusion

CI/CD Integration in SkillKit bridges the gap between manual skill authoring and automated production environments. By utilizing standardized templates for GitHub, GitLab, and local Git hooks, the system ensures that AI agent instructions are consistently validated and tested. This automation is critical for maintaining the reliability of "Verification Gates," allowing developers to catch errors early in the lifecycle and maintain a high standard for portable AI agent skills.

Sources: README.md:144-150, packages/core/src/methodology/packs/planning/verification-gates/SKILL.md:160-180