harness/harness-skills

Harness-skills is a tactical 'adapter' repo designed to ensure the Harness CI/CD platform remains accessible in the emerging 'agentic' developer workflow. With only 4 stars and 36 days of age, it is in its earliest stages. The defensibility score is low (3) because the code itself is largely a collection of schema definitions and wrappers for existing Harness APIs; there is no proprietary moat in the 'skills' logic itself. The real value and 'data gravity' reside in the Harness platform, which this project simply exposes to LLMs. Frontier labs (OpenAI/Anthropic) have no interest in building Harness-specific tools, making frontier risk low. However, platform risk is medium because as LLM context windows grow and coding agents become more capable of using raw documentation or OpenAPI specs to interact with any platform, the need for manually curated 'skills' may diminish. Competitors like GitHub (with native Copilot/Actions integration) and GitLab (Duo) pose a significant threat by offering a more vertically integrated experience that doesn't require third-party skill configuration. The project represents a necessary evolution for enterprise SaaS—providing the 'hooks' for AI agents to operate their software—but it remains a thin layer over a large legacy infrastructure.