name	cli-evals
description	Run, inspect, and debug cli-evals suites for live coding-agent CLI evaluations.

CLI Evals

Use this skill when a user wants to run a live eval suite, compare agents, inspect reports, rebuild metrics, or diagnose local prerequisites.

Run cli-evals doctor first when the target agent or auth state is unclear.
List scenarios with cli-evals list --suite <config>.
Run canaries before full corpora: cli-evals run --suite <config> --canary --agent <agent> --model <model>.
Use --filter <EV-ID> for a failing scenario and add --keep to preserve the workdir.
Rebuild summaries from existing reports with cli-evals rebuild --report <evals/reports/latest.json>.

Do not trigger --all live agent runs without the user explicitly asking; they cost time, tokens, and may use authenticated agent sessions.

name	cli-evals
description	Run, inspect, and debug cli-evals suites for live coding-agent CLI evaluations.

Use this skill when a user wants to run a live eval suite, compare agents, inspect reports, rebuild metrics, or diagnose local prerequisites.

Run cli-evals doctor first when the target agent or auth state is unclear.
List scenarios with cli-evals list --suite <config>.
Run canaries before full corpora: cli-evals run --suite <config> --canary --agent <agent> --model <model>.
Use --filter <EV-ID> for a failing scenario and add --keep to preserve the workdir.
Rebuild summaries from existing reports with cli-evals rebuild --report <evals/reports/latest.json>.

Do not trigger --all live agent runs without the user explicitly asking; they cost time, tokens, and may use authenticated agent sessions.

cli-evals