Coding Agent Skills

MassTransit publishes focused Markdown guidance for coding agents. The guidance covers common implementation tasks, uses MassTransit v9 syntax, and links to the canonical documentation for complete configuration details.

Use the Agent Index

Start an agent with the site’s LLM index. It is a concise map of the available skills and canonical documentation. Agents that support llms.txt can discover this conventional endpoint directly; for other agents, provide the URL as project context or open it before asking the agent to make a MassTransit change.

The index points to focused raw Markdown files under /skills/. Choose the skill that matches the change instead of loading every skill:

Task	Skill
Consumer or consumer definition	Consumer Foundation
Saga state machine, persistence, or scheduled event	Saga State Machine Foundation
Routing slip activity or compensation	Routing Slip Activity Foundation
Request client or response consumer	Request Response Foundation
Test harness and message-flow tests	Test Harness Foundation
Retry, redelivery, or an outbox	Outbox Foundation
Endpoint naming, concurrency, or middleware	Endpoint Configuration Foundation

The skills runbook explains how to combine skills for a larger change.

Agent Guidance

Skills are implementation guidance, not a substitute for the target application’s code or package version. An agent should:

1. Read the relevant skill and linked canonical documentation.
2. Preserve intentional conventions in the target application.
3. Verify APIs against the MassTransit version used by the application.
4. Add or update focused tests, then run the application’s build and test commands.

Message contract classes, records, and interfaces are all supported by MassTransit. The skills prefer records for new application-owned contracts, but agents should retain an existing contract shape when interoperability or versioning requires it.

Standards and Compatibility

/llms.txt is the emerging, widely adopted convention for an agent-readable website index. It is a proposal rather than an IETF web standard, so agent support varies. It improves discovery for agents that browse documentation, but it does not automatically install instructions into a local coding-agent session and it does not grant permission for model training or crawling. Those policies remain controlled by the site’s normal access and robots.txt settings.

The Agent Skills specification serves a different purpose: it defines a local skill package as a directory containing a SKILL.md file with YAML metadata. The raw /skills/*.md files are web-readable reference material, not a packaged Agent Skills distribution. Tools that use the Agent Skills format can download or vendor the selected guidance into their own skill package, while web-browsing agents should use /llms.txt.

For raw-skill clients, /skills/llms.txt provides a scoped index for that directory.