← ClaudeAtlas

usagelisted

Token usage and API-equivalent cost of this conversation or all of them: tokens, cache, context fill, per-turn breakdown, subagents, plan limits, plus a live side panel. Use when the user asks about tokens, context, cost or limits («сколько потратили», «сколько осталось лимита», /usage, /cost), and on your own when a long conversation may be near its context limit, to decide whether to suggest /handoff.
stxrmik/antigravity-plus · ★ 1 · AI & Automation · score 74
Install: claude install-skill stxrmik/antigravity-plus
# /usage — token & cost report > **On your own:** in a long conversation, check the context fill (`usage.py --conversation <id>`, read only the summary). Say nothing unless it's above ~70% (or ~170k+ on Gemini, where Antigravity's limit is ~300k and compaction triggers around ~250k); then tell the user in one line and suggest /handoff. The script [usage.py](scripts/usage.py) reads Antigravity transcripts (each model call records `input_tokens`, `cache_read_tokens`, `output_tokens`) and prices them with [pricing.json](pricing.json). ## Steps 1. **Pick the mode from the user's request:** - default → the current session. Your conversation id is in your system context (also the `<id>` segment of your artifact directory path). - "all", "total", "за неделю", "по дням" → `--all` (add `--days N` if a period is mentioned). - If the user names a model to price against ("посчитай как Sonnet"), add `--model <id>` using an `id` from pricing.json. - If the user names a Gemini API tier ("по Batch", "Flex", "Priority", "Contributor"), add `--tier <name>` (lower-case); models without that tier in pricing.json stay on Standard. 2. **Run it** (sandboxed first): ```bash python3 <skill_dir>/scripts/usage.py --conversation <conversation-id> --html <artifact_dir>/usage.html ``` For all sessions: ```bash python3 <skill_dir>/scripts/usage.py --all --days 7 ``` - `--all` and subagent transcripts need read access to the whole `~/.gemini/antigravity/brain/`