This walks you through the full lifecycle of adding function calling to AI agents in LaunchDarkly: identifying capabilities, creating tool schemas, attaching them to config variations, and verifying everything works. It's opinionated about workflow (always list, create, attach, verify, don't stop halfway) and includes conversion snippets for every major provider since LaunchDarkly stores a flat schema but OpenAI wants a nested function wrapper, Anthropic wants input_schema instead of parameters, and so on. The documentation is unusually detailed about a UI bug where attaching tools can clobber other fields, and explicitly tells you not to work around it by echoing old values back. Useful if you're managing tools across variations and need the guard rails.
npx skills add https://github.com/launchdarkly/agent-skills --skill aiconfig-tools