# Claude Code skill listing budget formula > Docs: the budget scales at 1% of the model's context window; on overflow, descriptions drop starting with the least-invoked skills; every name stays. Parent: [Mac local LLMs: Agent clients, context and compaction](https://llms-explorer.com/tree/mac-local-llms-agent-clients-context-and-compaction/) · 1 facets · 5 facts · page: https://llms-explorer.com/tree/claude-code-skill-listing-budget-formula/ ## Facts - Docs: the budget scales at 1% of the model's context window; on overflow, descriptions drop starting with the least-invoked skills; every name stays. — source: `asserted` - The skills page says the Skills row in /context reports the listing after budget; a debug-log warning appears on overflow (existing dossier). — source: `asserted` - Exact formula, including how tokens convert to characters and why a name-only floor exists. — source: `asserted` - The skills docs state only the 1% of context window rule, the skillListingBudgetFraction setting, the SLASH_COMMAND_TOOL_CHAR_BUDGET override and the 1,536-character per-entry cap; they give no characters-per-token factor or minimum — [source](https://code.claude.com/docs/en/skills) - This claim set adds nothing beyond agent-skill-listing-injection-cost.md — source: `asserted`