Problem
Skills in this plugin have no stability indicator. An AI agent consuming skills cannot distinguish a well-tested default skill from a new/beta one, causing it to treat experimental workflows as production-ready. This leads to wasted tool calls on deprecated MCP servers or experimental features whose APIs change without notice.
Proposed solution
Add a stability: frontmatter field to the SKILL.md schema with three values:
experimental — actively developed, may change without noticestable — tested, production-ready, backwards-compatibledeprecated — maintained for compatibility but no longer actively developed
Skill authors set stability: in frontmatter when submitting. The plugin's README or skill discovery surface can document the field so agents can filter or warn accordingly.
Use case
An AI agent scanning a large plugin pool for production-grade workflows can filter on stability: stable and warn before invoking experimental skills or skip deprecated ones entirely. This prevents wasted tool calls and incorrect routing assumptions.
Problem
Skills in this plugin have no stability indicator. An AI agent consuming skills cannot distinguish a well-tested default skill from a new/beta one, causing it to treat experimental workflows as production-ready. This leads to wasted tool calls on deprecated MCP servers or experimental features whose APIs change without notice.
Proposed solution
Add a
stability:frontmatter field to the SKILL.md schema with three values:experimental— actively developed, may change without noticestable— tested, production-ready, backwards-compatibledeprecated— maintained for compatibility but no longer actively developedSkill authors set
stability:in frontmatter when submitting. The plugin's README or skill discovery surface can document the field so agents can filter or warn accordingly.Use case
An AI agent scanning a large plugin pool for production-grade workflows can filter on
stability: stableand warn before invokingexperimentalskills or skipdeprecatedones entirely. This prevents wasted tool calls and incorrect routing assumptions.