I use it in interactive Claude Code sessions. For example, if I want to add a Claude API feature to an app, I'll often hack it together first in CC. This skill lets me start and stop tracking so I get get a sense for what the API costs will be.
For example:
> /cost-tracking > [do stuff] > stop tracking
Then I get a nice little table showing tokens and cost for the model that was used:
Haiku cost report (this session's work):
┌───────────────────┬───────────┬─────────┐
│ Category │ Tokens │ Cost │
├───────────────────┼───────────┼─────────┤
│ Base input │ 182 │ $0.0002 │
├───────────────────┼───────────┼─────────┤
│ Output │ 3,276 │ $0.0164 │
├───────────────────┼───────────┼─────────┤
│ Cache writes (1h) │ 25,619 │ $0.0512 │
├───────────────────┼───────────┼─────────┤
│ Cache reads │ 1,585,531 │ $0.1586 │
├───────────────────┼───────────┼─────────┤
│ Grand Total │ │ $0.2264 │
└───────────────────┴───────────┴─────────┘No comments yet.