Extending Coro
Coro’s TypeScript stays boring on purpose — intelligence lives in Markdown workflows while plugins supply provider-specific behaviour (SCM, trackers, LLMs). This section is for engineers packaging new integrations or hardening internal guardrails.
Use it after skimming Concepts and the Plugin SDK overview.
| Page | What you will learn |
|---|---|
| Plugin manifest | PluginManifest fields: validation schema, webhooks, MCP maps, intelligence hooks |
| SCM plugins | ScmPluginBase lifecycle + MCP-first pattern exemplified by @coro-ai/plugin-gitlab |
| Tracker plugins | Native vs MCP modes for getIssue / commentIssue / transitionIssue |
Executor plugins | Building a PhaseExecutor (Anthropic vs OpenAI implementations) |
| Custom guardrails | Scriptable rules + JSON overrides |
| Intelligence snippets | Packaging .md hints layered into every job |
| Webhooks | PluginManifest.webhook, HMAC helpers, normalizeInbound |
| Conformance testing | Harness expectations for third-party plugins |
| Publishing plugins | coro plugin init, npm publish, install trust model |
Repository source of truth: coro-developer-framework (packages/plugin-sdk, packages/runner, packages/plugin-gitlab, packages/llm-openai, …).