analyze-test-coverage

Analyze Test Coverage

Given a set of code changes, produce a structured test coverage analysis that answers three questions:

1. What tests were introduced or modified? (and what behavior does each one verify)
2. Is every behavior change covered by a test? (change coverage)
3. Is there code that no test exercises at all? (uncovered code)

Changing behavior without a test is a reliable way to lose that behavior in a future refactor. This skill surfaces those risks explicitly.

Prerequisites

• A git worktree or repository with the changes to analyze
• If called by a parent skill, $WORKTREE_DIR may be set and the diff context may be provided
• If invoked directly, $1 is either a diff ref (e.g. origin/main..HEAD) or a worktree directory

Inputs

The skill accepts its context in one of these forms:

Source	How
Worktree + diff ref	$WORKTREE_DIR set by parent skill; diff is git diff against the base branch
Direct invocation with diff ref	$1 is a git ref range (e.g. origin/main..HEAD); run in the current repo
Direct invocation with worktree	$1 is a directory path; diff is origin/<base>..HEAD inside that worktree

From the parent skill, the following may be available:

• CHANGED_FILES: list of files changed in the diff
• DIFF: the full diff text
• HEAD_COMMIT / SHORT_SHA: the commit being reviewed

If not provided, derive them:

if [ -n "${WORKTREE_DIR:-}" ]; then
  cd "$WORKTREE_DIR"
fi
DIFF_REF="${1:-origin/main..HEAD}"
git diff --name-only "$DIFF_REF"
git diff "$DIFF_REF"

Steps

1. Separate test files from source files

From the changed files list, split into:

• Test files: files matching the project's test conventions (test_*.py, *_test.py, *.test.ts, *.spec.ts, *_test.go, etc.)
• Source files: all other changed files (excluding config, docs, lock files, etc.)

# Identify test files by convention
git diff --name-only "$DIFF_REF" | grep -E '(test_|_test\.|\.test\.|\.spec\.|_test\.go|tests/)' || true