// Load when writing or reviewing skills. Design guidance for SKILL.md files and bundled resources.
| name | skill-artifacts |
| type | reference |
| description | Load when writing or reviewing skills. Design guidance for SKILL.md files and bundled resources. |
| model-invocable | false |
Skills are reference material loaded into agents. Each skill is a directory
under skills/ with a SKILL.md and optional resources/, scripts/,
and assets/ subdirectories.
Descriptions lead with when to load the skill, not what it contains. Most
skills should set model-invocable: false — only exempt skills that serve as
safety nets an agent might need to discover mid-task.
Set type: to control injection order — principle loads first (highest
attention), guardrail second (safety boundaries), reference last
(operational how-to). Omit for skills where order doesn't matter.
Keep the body under 500 lines. Put deep or variant-specific content in
resources/ and route to it from the body.
Load when writing or reviewing agent .md files. Design guidance for agent definitions and meridian-specific conventions.
Load when writing, reviewing, or improving any prompt artifact. Research-backed principles across four levels: prompt craft, skill extraction, agent design, and multi-agent coordination.
Load when reviewing agent definitions or skill content. Adversarial review methodology — finding quality, severity, and structured reporting.
