What does this skill do, and when should you use it?
Swift Concurrency Pro is an Agent Skill by Paul Hudson (author of Hacking with Swift) that reviews Swift code for concurrency correctness, modern API usage, and common async/await pitfalls. It ships with twelve reference documents covering actor reentrancy, cancellation, AsyncStream, GCD migration, and Swift 6.2 changes. The skill targets projects on Swift 6.2 or later with strict concurrency checking enabled. The author is explicit that it is not a dump of all concurrency knowledge — it focuses only on rules that change LLM behavior. It is one of two skills bundled in the repository and is MIT licensed.
- Runs a 12-step review flow: scans dangerous-pattern hotspots first, then checks actor isolation, structured concurrency, cancellation, AsyncStream, sync/async bridging, and more
- Reports each issue with file, line numbers, the violated rule, and a before/after code fix
- Ends with a prioritized summary of the most impactful changes
- Supports partial reviews by loading only the relevant reference files
- Actively blocks common bad advice, such as using @unchecked Sendable to silence compiler errors or adding third-party concurrency frameworks without asking
- iOS/macOS developers migrating a GCD- or completion-handler-based codebase to async/await who want races and broken cancellation caught
- Teams facing a wall of Swift 6.2 strict-concurrency compiler errors who need diagnostics mapped to correct fixes
- Engineers using Claude Code or Codex for pre-merge review who want actor reentrancy and unstructured-tasks-in-loops flagged
- Reviewers of Swift concurrency code who want focused checks on AsyncStream continuation lifecycle and back-pressure
- Maintainers of mixed TaskGroup / Task {} codebases verifying structured concurrency is used where it should be
- Projects targeting non-Swift languages or platforms — the entire skill is Swift concurrency specific
- Projects pinned to pre-6.2 toolchains with no plan to upgrade — the skill assumes Swift 6.2 strict concurrency
- Beginners looking to learn async/await basics — the author deliberately avoids reteaching fundamentals models already know
How do you install this skill?
- This is a static source-only review; no skill instructions were executed and confidence is low.
- openai.yaml enables implicit invocation, which may consume tokens when not needed; disable it if you want explicit control.
- The skill targets Swift 6.2 strict concurrency; advice may not fit older projects — verify your target environment.
- No test suite or changelog; reference correctness is not independently execution-verified, so review findings still need human confirmation.
- Publisher is unverified in the FollowSkills registry; identity is treated as unknown, not as a deduction or suspicion.
- Local filesystem
Claude Code: add the skill marketplace, then install the plugin.
/plugin marketplace add twostraws/Swift-Concurrency-Agent-Skill
/plugin install swift-concurrency-pro@swift-concurrency-agent-skillCodex, Gemini, Cursor, and others: install via npx, selecting which agents and which scope during installation.
npx skills add https://github.com/twostraws/swift-concurrency-agent-skill --skill swift-concurrency-proIf you see npx: command not found, install Node via Homebrew first (brew install node). You can alternatively clone the whole repository and install it however you want; Xcode users can follow the author's walkthrough video.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- /swift-concurrency-pro Make sure my project takes full advantage of Swift concurrency
- $swift-concurrency-pro Help me migrate this project to task groups
- Use the Swift Concurrency Pro skill to look for potential concurrency problems in this project
- Review DataLoader.swift, focusing on actor reentrancy and cancellation propagation
Trigger via slash command or natural language: /swift-concurrency-pro in Claude Code, $swift-concurrency-pro in Codex. Append instructions for a partial review (e.g., only checking task-group migration). Once triggered, the skill follows the 12-step flow in SKILL.md: it uses references/hotspots.md to prioritize inspection, then works through actor, structured/unstructured concurrency, cancellation, AsyncStream, bridging, and GCD-migration reference docs, ending with a per-file issue list and a prioritized summary.
What are this skill's strengths and limitations?
- Authored by someone who maintains a Swift concurrency book and Swift 6.x release articles; rules target real LLM failure modes, not generic advice
- Twelve focused reference documents cover Swift 6.2 features (default actor isolation, @concurrent, Task.immediate, etc.) that models weren't trained on
- Well-defined output format: per-file issues with before/after fixes and a prioritized summary
- MIT licensed for commercial use, modification, and distribution; explicitly bans dangerous advice like @unchecked Sendable as a quick fix
- Marked version 1.0; the repository provides no test suite or effectiveness benchmarks, so interception quality is unverifiable from the source
- Assumes Swift 6.2+ strict concurrency, limiting fit for projects on older toolchains
- The repository bundles 2 skills, so installing the whole collection brings in unrelated content
- Pure text rules and reference docs — no executable scripts or automated checking tooling
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Swift Concurrency Pro this page | 65 · Recommended | ★ 572 | 5mo ago | MIT |
| Swift Concurrency Pro | 60 · Recommended | ★ 572 | 5mo ago | MIT |
| SwiftUI Pro | 57 · Use with care | ★ 5.9k | 18d ago | MIT |
| SwiftUI Pro | 51 · Use with care | ★ 5.9k | 18d ago | MIT |
| cmux Architecture Skill | 63 · Recommended | ★ 28k | 11d ago | NOASSERTION |
The README points to the author's companion skills (SwiftUI Pro, SwiftData Pro, Swift Testing Pro) as complementary picks, but the source contains no directly comparable competing product for this skill.
How did FollowSkills review this skill?
Pure reference-documentation skill: no scripts, no network calls, no file writes or external side effects; near-minimal privilege. MIT license and clear author/repository attribution; fully auditable source. Deductions: publisher not verified (no deduction for that alone per rules, but rollback/install-change notes only partially covered), and openai.yaml sets allow_implicit_invocation: true, removing an explicit user-confirmation gate.
Instructions are self-consistent across SKILL.md and references (reentrancy and @unchecked Sendable guidance cross-referenced consistently); all 12 referenced files match the declared workflow. Static cap of 10 applies: no test suite, no CI execution evidence, no runnable key path; failure-feedback quality depends on the executing LLM, not the skill itself.
Trigger description is clear (use when reading/writing/reviewing Swift concurrency code), audience and boundaries explicit (does not reteach LLM fundamentals, no third-party frameworks), partial review supported. Deductions: implicit invocation may cause false triggers on any Swift contact; no Chinese-language support statement. Purely local docs with no overseas-service dependency, so reachability is fine.
Well-layered docs: concise SKILL.md respecting token budgets, 12 progressive-disclosure reference files, README with install, usage, contributing, and license; author and maintenance responsibility clear (Paul Hudson, active maintainer). Deductions: version is only "1.0" with no changelog, limited known-limitations disclosure (no stated fallback for pre-6.2 projects), no troubleshooting section.
Output format is well defined (per-file, line numbers, rule names, before/after diffs, prioritized summary); examples are concrete and usable; clear marginal value targeting LLM-specific concurrency mistakes (reentrancy, double-resume, etc.). Static cap of 7: no verified real-project review output; correctness of Swift 6.2 behavior claims unverified by execution; benefit must be confirmed in use.
Primary material is auditable: reference files contain checkable code examples and external links (massicotte.org, Apple APIs), with facts and recommendations reasonably separated. Capped at 5: no test reproduction, no third-party execution evidence; key claims about Swift 6.2 behavior are supported only by the author's own articles, without independent corroboration.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →