An agent skill for Apple's Swift Testing framework (import Testing), covering Swift 6.0-6.4 and Xcode 16-27. It works with any agent that reads the Agent Skills format, including Claude Code and OpenAI Codex.
It focuses on what coding agents get wrong: tests that compile but pass when they should fail (Task { } in a test, asserting inside a callback, XCTAssert* inside @Test), the scope of .serialized, invented APIs (a .repeating trait, sub-minute time limits), and Swift 6.1-6.4 additions such as exit tests, image and Transferable attachments, Test.cancel, issue severity, and CustomTestReflectable. For XCTest migrations it includes a compact mapping and corrects consequential gaps in Apple's modernize-tests recipes.
The skill follows the repository's existing framework: it writes XCTest in XCTest targets unless asked to use Swift Testing or to migrate.
Codex and other agents that read ~/.agents/skills:
git clone https://github.com/farkasseb/swift-testing-skill ~/.agents/skills/swift-testingClaude Code reads ~/.claude/skills. Clone there, or symlink the copy above:
mkdir -p ~/.claude/skills
ln -s ~/.agents/skills/swift-testing ~/.claude/skills/swift-testingFor a single project, put it in the repository's .agents/skills/ (Codex) or .claude/skills/ (Claude Code).
SKILL.md Target settings, false-pass review, reference routing
references/modern-apis.md Errors, exit tests, attachments, severity, cancellation, scoping and issue-handling traits
references/concurrency-and-structure.md Suites and lifecycle, parallelism, confirmation, time limits, skipping, tags, parameterized tests
references/migration.md XCTest migration beyond the assertion mapping, XCTest interop, test plans, verification
evals/evals.json Focused prompts with review expectations
tests/run.sh Verification harness (see below)
tests/run.sh runs on the installed release baseline, Xcode 26.6 / Swift 6.3.3, and Xcode 27.1 beta / Swift 6.4. Swift 6.4-only checks are explicitly skipped on 6.3. Select Xcode with $XCODE (an .app path), $DEVELOPER_DIR, or xcode-select -p:
XCODE=/Applications/Xcode-26.6.0.app TEST_LOG_DIR=/tmp/testing-release tests/run.sh
XCODE=/Applications/Xcode-27.1.0-Beta.app TEST_LOG_DIR=/tmp/testing-beta tests/run.shThe suite compiles and runs the executable examples, checks callback timeout/cancellation and per-case scoping, and verifies package interoperability using process status, the named test outcome and runtime warnings. It also checks API/deployment availability in the installed SDK. Some snippets only check compilation; empty example bodies are not evidence of behavior.
The October 2026 audit additionally ran deliberately broken async/callback tests, shared-state suites, parameterized/scoped tests, and iOS simulator interop. Evidence distinguishes initial false passes, actual assertion failures, compiler errors and infrastructure/watchdog failures. A green suite does not prove every possible migration preserves behavior.
Audited libraries: Xcode 26.6 (Swift 6.3.3, Testing 1902) as the release baseline; Xcode 27.1 beta (Swift 6.4, Testing 2084) for the main audit. Xcode 27.0 RC has the same public Testing interfaces. Xcode 27.2 beta 2 reports Testing 2401 with the same public declarations.
evals/evals.json holds prompts with graded expectations. Most target knowledge that models trained before the 6.3/6.4 releases get confidently wrong; others check migration correctness, behavior preservation, and the keep-XCTest boundary. Run each prompt with and without the skill, with web and documentation lookup disabled, so the difference measures the skill.
- Swift Testing documentation and swift-testing on GitHub
- Swift Testing evolution proposals, ST-0001 through ST-0028
- Meet Swift Testing and Go further with Swift Testing (WWDC24)
PRs are welcome. Include a source (Apple documentation, an evolution proposal, or a reproducible test) for any behavioral claim.