GPT-4o mini
- Total startup tokens
- 0
- Total triggered tokens
- 0
- Startup context share
- Unknown
- Startup / 1,000 sessions
- Price unknown
Skill Token Cost Calculator
Measure startup and triggered skill tokens, context share, and estimated cost for 1,000 sessions across models.
Step 1 · Skills & models
1 / 50 skillsAdd skills, then choose comparison models
Add multiple files over time; every change recalculates the results below.
Skill sources
394 B- release-notes.md
Target models
1 / 5GPT-4o mini
Up to 5 models. Unknown prices are never treated as free.
Startup = frontmatter name + description; triggered = complete body. Cost estimates use input prices for 1,000 sessions and depend on the host. Combined ≤5 MB, each file ≤512 KB.
Step 2 · Compare results
Token & cost results
Add skills and select models to see results.
Made for the task at hand.
Estimate how a set of Agent Skills affects context and input-token spend. Startup usage counts frontmatter name plus description; triggered usage counts the complete SKILL.md body. The existing Tokenizer Worker and model snapshot provide encoding, context-window, and price data without duplicating the catalog.
How to use
- Add up to 50 SKILL.md files by paste, multi-file selection, or explicit direct URLs, keeping the combined input within 5 MB.
- Select one or more target models and review per-skill startup and triggered token counts.
- Compare context share and estimated input cost per 1,000 sessions, then copy or download the complete CSV.
Your work stays yours.
Tool inputs are processed locally in your browser and are never sent to KitVerse servers. Clear the workspace when using a shared device.
Tool inputs are processed locally in your browser and are never sent to KitVerse servers. Clear the workspace when using a shared device.
Common questions
Are these the exact costs charged by an Agent host?
No. Counts use the selected model encoding and public build-time price snapshot. Actual loading, caching, wrappers, and billing depend on the host. Models without compatible mappings are marked as estimates.
What is included in each token count?
Startup is the frontmatter name plus description only. Triggered usage is the full body after frontmatter, not the entire file. Cost estimates use input price per million tokens for 1,000 sessions; unknown prices remain unknown rather than zero.