# Test Gap

> Find changed branches that no test meaningfully asserts, ranked by the consequence if they are wrong. Use explicitly when the user asks to review a diff for coverage, identify missing tests before merging, or inspect test gaps in a branch or package. Do not write tests unless asked afterward.

- Skill: `handsomefox/test-gap` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add handsomefox/test-gap`
- Raw SKILL.md: https://api.skillmd.com/api/skills/handsomefox/test-gap/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: handsomefox (https://skillmd.com/u/handsomefox)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/handsomefox/test-gap

---


# Test gap

Scope: staged changes by default, or a named branch or package. Nothing outside
it appears in the output.

## Measure

Run the repo's coverage tooling and read the data. **Never estimate coverage by
reading source.** If coverage tooling is not configured here, say so and stop.

## Cross-reference

Intersect coverage with the diff hunks. A line executed by an unrelated test is
executed, not tested — check that some test asserts on that branch's behavior
rather than merely reaching it. Report those as "executed, unasserted."

## Rank by consequence

Silently wrong results, then data loss, then crashes, then degradation. Error
paths, boundaries, and concurrent access before happy paths.

## Exclude

Generated code, trivial accessors, anything the diff did not touch, and tests
that would assert the implementation matches itself.

## Output

Table: location, uncovered branch, consequence if wrong, nearest existing test,
the assertion that closes it.