Context
The harness can run one scenario at a time. Teams will quickly want a suite mode that runs a directory of scenarios against corresponding traces or targets and produces one aggregate result.
Acceptance criteria
- Design a minimal suite command or flag that keeps single-scenario behavior intact.
- Support trace-file based suites first.
- Emit per-scenario result JSON and an aggregate summary.
- Exit behavior should compose with the CI fail-on-result option.
- Add tests for partial failures and missing trace mappings.
- Document expected directory conventions.
Useful files
src/agent_harness/cli.py
src/agent_harness/runner.py
tests/test_cli.py
docs/ci-github-actions.md
Context
The harness can run one scenario at a time. Teams will quickly want a suite mode that runs a directory of scenarios against corresponding traces or targets and produces one aggregate result.
Acceptance criteria
Useful files
src/agent_harness/cli.pysrc/agent_harness/runner.pytests/test_cli.pydocs/ci-github-actions.md