Measures and reports token usage, elapsed time, and cost for plugin skill runs.
npx skills add https://github.com/cagriy/dev-skills --skill usage-report-cagriy
Please help me install this Agent Skill. Skill: usage-report Source: https://github.com/cagriy/dev-skills/tree/main/skills/usage-report Command: npx skills add https://github.com/cagriy/dev-skills --skill usage-report-cagriy
start marker records when a run began, and report measures the window from that marker to the present moment. - Detailed usage table: Emits a markdown table with elapsed time, run time, token totals split across main agent and subagents, request counts, and output throughput. - Durable logging and tracker integration: Appends one line to a persistent usage log and writes the figures into the feature tracker's stage panel via anchored HTML tokens. - Failure-safe design: Every problem degrades to a single usage report skipped line, so reporting never fails, blocks, or aborts the calling skill. - Use Case: A feature-implementation run finishes and the user sees exactly how many tokens the main agent and its eval subagents consumed, how long it took, and the throughput, all logged for later comparison across features. ## Quick Start This Skill is invoked internally by other plugin skills with an argument like report feature-implement, tracker_file=<path>, outcome=completed and is not called directly by users.High-intent search queries and answers about installing and using this skill.
Invoke the usage-report skill with `start <slug>` at the beginning of a run and `report <slug>` at the end. The report mode measures the window between the two markers and prints a markdown table with token totals, request counts, and throughput.
Each report run appends one line to a durable usage log keyed by session id and skill slug, so parallel sessions never collide. The log accumulates per-run figures that can be compared across features over time.
No, the skill is marked user-invocable: false and is only called internally by the four feature-chain skills. It takes its pathing as input from the caller and never asks the user for anything.
The skill prints a single `usage report skipped` line naming the cause and continues the caller's workflow. It never guesses a start time, never retries, and never blocks or aborts the calling skill.
Yes, the underlying script scans the session transcript and splits the token breakdown into main agent, subagents, and total. Subagents never call the skill themselves; their work inside the window is counted by the scan.
yimwooCoordinates HOTL skill-driven workflows for implementation tasks from brainstorm to execute stages.
droxeyEstimate monthly model costs from token usage and pricing inputs.
AradotsoRender Markdown files in a live-updating panel beside the terminal.
arndvsAnalyzes agent session transcripts to record measured codebase friction observations.