| name | agent-artifacts |
| type | reference |
| description | Load when writing or reviewing agent .md files. Covers agent artifact shape, frontmatter, and Meridian-specific agent conventions.
|
| model-invocable | false |
Agent Artifacts
Use /prompt-principles for broader prompt design guidance. This skill covers how agent artifacts are structured in Meridian.
Agent definitions are markdown files with YAML frontmatter. The body below the frontmatter is the system prompt. For the full frontmatter schema, see the mars agent profile reference.
Descriptions serve callers: they should help someone decide when to use the agent and what to pass. Bodies should stay caller-agnostic so the agent works from whatever context it receives, and harness-agnostic so they survive a harness swap — state intent, and leave CLI flags and tool mechanics to the environment (see /prompt-principles).
Managers and leads coordinate through spawns. They should not implement directly.
Load resources/meridian.md for Meridian-specific subagent, handoff, staffing, and package-target conventions. Load resources/model-policies.md for fallback patterns that keep agents usable across different model availability.
