SKM-P07 · Portability
`description` is longer than the tightest runtime budget
1,024 characters is the spec ceiling, but OpenClaw's own authoring guidance asks for under 160 and practical Claude Code reports suggest under 200 — a description written for the tightest constraint is safe everywhere.
What the rule is
The specification permits up to 1,024 characters, but every runtime with a published index budget is tighter than that in practice: OpenClaw's own authoring guidance asks for descriptions under roughly 160 characters, and a verified report from a Claude Code user with nine plugin skills found only about two reaching chat, with headroom recovered by keeping descriptions under about 200 characters.
Cursor and Hermes Agent publish no index-budget figure at all, which doesn't mean there isn't a cost — it means this tool can't state a specific number for either.
Where it comes from
- Agent Skills compatibilityRuntime documentation
- anthropics/claude-code #40774 — plugin skills excluded past the index budgetVerified issue — closed
Which runtimes enforce it
| Runtime | Effect | Note |
|---|---|---|
| Claude Code | Degrades | ~1% of context for the whole skill index; a real report saw about two of nine plugin skills reach chat past this budget. |
| Codex | Degrades | 2% of the context window or 8,000 characters, whichever is smaller, shared across the whole skill index. |
| Cursor | Not documented | No index budget is published for Cursor. |
| OpenClaw | Degrades | ~97 characters per skill plus field lengths against a configurable ceiling; degrades by shortening then dropping descriptions past it. |
| Hermes Agent | Not documented | No index budget is published for Hermes Agent. |
How to fix it
- Write the description for the tightest constraint rather than the spec's ceiling — under roughly 160 characters, trigger-focused, with the most important terms first. A description written this way is safe on every runtime this site has verified.
Elsewhere
- SKILL.md, the formatThe full frontmatter field table this rule is drawn from.
- Agent Skills Not WorkingThe diagnostic reference this rule’s portability and failure evidence comes from.
- All rulesEvery check the validator runs, grouped by class.