Skip to content

Harness and Usage Reports

VibePro records how agent workflows behave so repeated failures can become reviewable checks or learning candidates.

Harness status

bash
vibepro harness status .
vibepro harness map .

Use these commands to inspect whether agent workflow evidence is present and which surfaces are covered.

Record learnings

When a repeated failure pattern is clear, record it explicitly:

bash
vibepro harness learn . \
  --summary "<what failed repeatedly>" \
  --kind workflow \
  --source "<where it was observed>" \
  --pattern "<trigger pattern>" \
  --target review-gate

vibepro harness review-learnings .

Learning records should be concrete. Avoid recording vague advice that cannot become a future check, prompt, or skill.

Usage report

Usage reports summarize agent and subagent activity:

bash
vibepro usage report . --since 2026-06-01 --subagent-roi

Use them to understand whether additional agent review is worth its cost, where evidence is reused, and which manual interventions repeat often.

Released under the Apache-2.0 License.