| name | development-discipline |
| description | Developer-agent discipline for test judgment, validation evidence, and TDD-required task escalation. |
| version | 1.0.0 |
| category | core |
| triggers | development discipline, developer validation, test judgment |
| metadata | {"gobby":{"audience":"agent","depth":1}} |
Development Discipline
Use this for every developer-agent task.
Required Judgment
Before editing, make an explicit test judgment: decide how behavior can be
verified. Name the relevant test level: unit, integration, CLI/API, browser/UI,
migration/storage, or documented manual check.
Every completion or review handoff must include:
- Exact validation commands run.
- Result of each command.
- Why those commands cover the changed behavior.
- Any test gap that remains, with the reason it is acceptable.
Test Changes
Add or update tests when behavior changes. Skipping tests is acceptable only for
pure documentation, mechanical metadata, or code paths that cannot be executed
locally; document the reason in the task handoff.
When adding or heavily editing tests, run gobby test-quality audit on touched
test paths. For noisy areas, use:
uv run gobby test-quality audit <paths> --baseline .gobby/test-quality-baseline.json --fail-on-new --min-severity high
TDD-Required Tasks
If the task has label tdd:required, additional_skills contains
test-driven-development, or validation criteria require TDD evidence, load
test-driven-development before implementation and follow it exactly.