Thread Skill — How to Work#

Use this skill when the user asks about threads or concurrency: starting threads, thread pools, locks/mutex, race conditions, thread-safe code, or choosing between threads / processes / async.

Your Workflow#

1. Clarify context: Identify language/runtime (Python, Java, C#, Node.js, Go, etc.) and what they want (I/O-bound vs CPU-bound, parallelism vs concurrency).

2. Choose the right tool: Prefer the language’s recommended approach (e.g. async in Node/Python for I/O; threads or processes for CPU; see reference/concepts.md).

3. Give the right level of detail:

   • Quick answer: Use the Quick Reference below and reply with a minimal example in their language.
   • Patterns and pitfalls: Point to or quote reference/patterns.md (locks, queues, thread pool, shared state).
   • Language API: Point to reference/languages.md or write a short snippet following that reference.
   • Templates: Use or adapt code in assets/ when the user wants a starter or copy-paste pattern.

Quick Reference#

GoalPreferAvoidI/O-bound (network, disk)Async (async/await, asyncio, Promise) or thread poolMany blocking threadsCPU-bound (compute)Processes / process pool, or native threads (Java, Go, C#)Many threads in Python (GIL)Shared mutable stateLocks (mutex), thread-safe collections, or avoid sharingNaked shared variablesProducer–consumerQueue (thread-safe)Busy-wait / ad-hoc signalingRun N tasks in parallelThread pool or process poolManually creating N threadsConceptOne-linerRace conditionTwo threads read/write same data without synchronization; fix with lock or immutable design.DeadlockTwo (or more) threads wait for each other’s lock; fix by consistent lock order or timeouts.GIL (Python)Only one thread runs Python bytecode at a time; use processes or C extensions for CPU-bound.

Where to Look#

NeedLocationWhen to use threads vs processes vs asyncreference/concepts.mdLocks, queues, thread pool, shared state patternsreference/patterns.mdPython / Java / C# / Node / Go API quick refreference/languages.mdStarter code / templatesassets/

Safety and Best Practices#

• Do not suggest shared mutable state without synchronization (lock, queue, or thread-safe structure).
• Prefer higher-level constructs: thread pool, queue, async, instead of raw threads where they fit.
• For Python CPU-bound work, suggest multiprocessing or clarify GIL limits; for I/O-bound, suggest threading or asyncio as appropriate.
• When the user reports deadlock or race, point to reference/patterns.md and suggest minimal repro + lock discipline or queue-based design.

Assets#

• assets/python-thread-starter.py — Minimal Python threading + lock example.
• assets/python-thread-pool.py — Thread pool (e.g. concurrent.futures) example.

Add or reference other language templates in assets/ when the user asks for a specific runtime.