docs: add guidance on delegating failing tests to subagents #10

Merged

addison merged 1 commit from exe-dev-bot/shelley-extras:docs/subagent-failing-tests into main

2026-01-30 19:45:01 -05:00

exe-dev-bot commented

2026-01-30 19:42:13 -05:00

Contributor

Learning

When a test fails unexpectedly, the right response is to spawn a subagent to investigate and fix it - not to skip the test and move on.

Context

During a session adding DAO tests for Karriba, one test failed (importing from older schema versions). I lazily skipped it with a known bug comment instead of investigating.

The user correctly called this out as poor manager skills. When I spawned a subagent to investigate, it quickly found and fixed the actual bug in the migration logic (took about 5 minutes).

Change

Added to Subagents as Expert Consultants section:

Delegate failing tests - when a test fails unexpectedly, spawn a subagent to investigate and fix it; do not skip tests without proper investigation

## Learning When a test fails unexpectedly, the right response is to spawn a subagent to investigate and fix it - not to skip the test and move on. ## Context During a session adding DAO tests for Karriba, one test failed (importing from older schema versions). I lazily skipped it with a known bug comment instead of investigating. The user correctly called this out as poor manager skills. When I spawned a subagent to investigate, it quickly found and fixed the actual bug in the migration logic (took about 5 minutes). ## Change Added to Subagents as Expert Consultants section: - **Delegate failing tests** - when a test fails unexpectedly, spawn a subagent to investigate and fix it; do not skip tests without proper investigation

exe-dev-bot added 1 commit

2026-01-30 19:42:14 -05:00

docs: add guidance on delegating failing tests to subagents cbda20135a

When a test fails unexpectedly, the right response is to spawn a
subagent to investigate and fix it - not to skip the test and move on.

This learning came from a session where I lazily skipped a failing
test instead of delegating the investigation to a subagent. The
subagent quickly found and fixed the underlying bug.

Co-authored-by: Shelley <shelley@exe.dev>