agentmark-ai

Build, debug, and ship AgentMark prompts, datasets, experiments, and evals. TRIGGER when: working with `.prompt.mdx` files, `agentmark.json`, `agentmark.client.ts`, `agentmark_client.py`, an `agentmark/` or `.agentmark/` directory, or imports from `@agentmark-ai/*`; user says any of "set up AgentMark in this project", "wire AgentMark into my app", "integrate AgentMark", "run an AgentMark experiment", "create an AgentMark prompt", "deploy an AgentMark prompt", or runs `npm create agentmark` / `npx create-agentmark`; user runs or asks about `agentmark <cmd>` / `npx agentmark <cmd>` (`dev`, `run-prompt`, `run-experiment`, `build`, `generate-types`, `generate-schema`, `link`, `login`, `logout`, `pull-models`); user asks about driving AgentMark Cloud programmatically (the `agentmark-mcp` server, MCP tools, headless app provisioning, git-based deploys); user mentions AgentMark or asks about prompt versioning, dataset experiments, prompt evaluations, prompt deployments, or trace observability in an AgentMark project