What problem does it solve?
When tracking AI usage costs, per-agent and per-model reports hide which individual conversations consumed the most budget. This Skill aggregates cost data by conversation session so you can identify expensive sessions and evaluate whether a project's session pattern is sustainable.
Core Features & Use Cases
- Per-Session Cost Table: Lists every conversation with started-at timestamp, session ID, message count, top model, and total cost.
- Flexible Output Formats: Emits a markdown table by default or JSON via the CONV_FORMAT environment variable, with CONV_LIMIT to restrict to the most recent N conversations.
- Per-Project Rollups: Supports namespace overrides via CONV_NAMESPACE to scope cost analysis to a specific project.
- Use Case: After a week of sessions, run the report to find which conversations ran long on Opus and decide whether to adjust model selection or session length.
Quick Start
Ask the assistant to run the cost-conversation report to show total spend per conversation session.