Guide engineers through adding and registering LLM providers to the anie harness.
npx skills add https://github.com/danielmriley/anie --skill adding-providers
Please help me install this Agent Skill. Skill: adding-providers Source: https://github.com/danielmriley/anie/tree/main/.claude/skills/adding-providers Command: npx skills add https://github.com/danielmriley/anie --skill adding-providers
The skill provides a structured pathway for teams to add, scaffold, plan, and register new LLM providers or provider families to the anie harness. It codifies the end-to-end process, including native (new ApiKind) and aggregator (OpenAI-compatible) paths, discovery and telemetry integration, and the validation invariants that ensure consistent behavior across providers.
Follow the six-step recipe to implement and register a new provider.
High-intent search queries and answers about installing and using this skill.
Adding a new LLM provider involves scaffolding provider traits, declaring capabilities, extending the ContentBlock structure, implementing round-trip contract rules, and validating through invariant and rollout tests.
Native provider integration requires declaring a new ApiKind and implementing per-provider streaming semantics, whereas aggregator integration reuses the OpenAI-compatible path with a minimal per-vendor capability mapping to reduce duplicated code.
You declare model capabilities by extending the ContentBlock structure and specifying per-provider traits, ensuring you implement round-trip contract rules to maintain consistent behavior across the new LLM integration.
Validating a new provider requires adding invariant tests to verify round-trip contract rules, registering the provider in onboarding and UI, and verifying ecosystem gates with workspace checks during rollout.
Yes, you can reuse the OpenAI-compatible aggregator path for non-native providers by mapping per-vendor capabilities and handling specific quirks as a minimal compat-blob, which reduces duplicated integration code.
Provider onboarding fails ecosystem gates when round-trip contract rules are violated or capabilities are misdeclared. You must pass invariant tests and workspace checks to verify consistent behavior across the harness.
CNifeQuery the models.dev catalog for model specs, pricing, and provider API details.
EvanZhang008Orchestrate Playwright, API, and unit tests to verify Walnut features end-to-end.
jacob-balslevDesign and validate inbound webhook handlers for signature, duplicate, and retry issues.
chiptoe-svgConfigure OpenAI Codex as a NanoClaw agent provider with flexible authentication.
