Merge pull request #223 from shimo4228/feat/skills/regex-vs-llm-structured-text

feat(skills): add regex-vs-llm-structured-text skill

.cursor/skills/cpp-coding-standards/SKILL.md

@@ -0,0 +1,722 @@
+---
+name: cpp-coding-standards
+description: C++ coding standards based on the C++ Core Guidelines (isocpp.github.io). Use when writing, reviewing, or refactoring C++ code to enforce modern, safe, and idiomatic practices.
+---
+
+# C++ Coding Standards (C++ Core Guidelines)
+
+Comprehensive coding standards for modern C++ (C++17/20/23) derived from the [C++ Core Guidelines](https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines). Enforces type safety, resource safety, immutability, and clarity.
+
+## When to Use
+
+- Writing new C++ code (classes, functions, templates)
+- Reviewing or refactoring existing C++ code
+- Making architectural decisions in C++ projects
+- Enforcing consistent style across a C++ codebase
+- Choosing between language features (e.g., `enum` vs `enum class`, raw pointer vs smart pointer)
+
+### When NOT to Use
+
+- Non-C++ projects
+- Legacy C codebases that cannot adopt modern C++ features
+- Embedded/bare-metal contexts where specific guidelines conflict with hardware constraints (adapt selectively)
+
+## Cross-Cutting Principles
+
+These themes recur across the entire guidelines and form the foundation:
+
+1. **RAII everywhere** (P.8, R.1, E.6, CP.20): Bind resource lifetime to object lifetime
+2. **Immutability by default** (P.10, Con.1-5, ES.25): Start with `const`/`constexpr`; mutability is the exception
+3. **Type safety** (P.4, I.4, ES.46-49, Enum.3): Use the type system to prevent errors at compile time
+4. **Express intent** (P.3, F.1, NL.1-2, T.10): Names, types, and concepts should communicate purpose
+5. **Minimize complexity** (F.2-3, ES.5, Per.4-5): Simple code is correct code
+6. **Value semantics over pointer semantics** (C.10, R.3-5, F.20, CP.31): Prefer returning by value and scoped objects
+
+## Philosophy & Interfaces (P.*, I.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **P.1** | Express ideas directly in code |
+| **P.3** | Express intent |
+| **P.4** | Ideally, a program should be statically type safe |
+| **P.5** | Prefer compile-time checking to run-time checking |
+| **P.8** | Don't leak any resources |
+| **P.10** | Prefer immutable data to mutable data |
+| **I.1** | Make interfaces explicit |
+| **I.2** | Avoid non-const global variables |
+| **I.4** | Make interfaces precisely and strongly typed |
+| **I.11** | Never transfer ownership by a raw pointer or reference |
+| **I.23** | Keep the number of function arguments low |
+
+### DO
+
+```cpp
+// P.10 + I.4: Immutable, strongly typed interface
+struct Temperature {
+    double kelvin;
+};
+
+Temperature boil(const Temperature& water);
+```
+
+### DON'T
+
+```cpp
+// Weak interface: unclear ownership, unclear units
+double boil(double* temp);
+
+// Non-const global variable
+int g_counter = 0;  // I.2 violation
+```
+
+## Functions (F.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **F.1** | Package meaningful operations as carefully named functions |
+| **F.2** | A function should perform a single logical operation |
+| **F.3** | Keep functions short and simple |
+| **F.4** | If a function might be evaluated at compile time, declare it `constexpr` |
+| **F.6** | If your function must not throw, declare it `noexcept` |
+| **F.8** | Prefer pure functions |
+| **F.16** | For "in" parameters, pass cheaply-copied types by value and others by `const&` |
+| **F.20** | For "out" values, prefer return values to output parameters |
+| **F.21** | To return multiple "out" values, prefer returning a struct |
+| **F.43** | Never return a pointer or reference to a local object |
+
+### Parameter Passing
+
+```cpp
+// F.16: Cheap types by value, others by const&
+void print(int x);                           // cheap: by value
+void analyze(const std::string& data);       // expensive: by const&
+void transform(std::string s);               // sink: by value (will move)
+
+// F.20 + F.21: Return values, not output parameters
+struct ParseResult {
+    std::string token;
+    int position;
+};
+
+ParseResult parse(std::string_view input);   // GOOD: return struct
+
+// BAD: output parameters
+void parse(std::string_view input,
+           std::string& token, int& pos);    // avoid this
+```
+
+### Pure Functions and constexpr
+
+```cpp
+// F.4 + F.8: Pure, constexpr where possible
+constexpr int factorial(int n) noexcept {
+    return (n <= 1) ? 1 : n * factorial(n - 1);
+}
+
+static_assert(factorial(5) == 120);
+```
+
+### Anti-Patterns
+
+- Returning `T&&` from functions (F.45)
+- Using `va_arg` / C-style variadics (F.55)
+- Capturing by reference in lambdas passed to other threads (F.53)
+- Returning `const T` which inhibits move semantics (F.49)
+
+## Classes & Class Hierarchies (C.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **C.2** | Use `class` if invariant exists; `struct` if data members vary independently |
+| **C.9** | Minimize exposure of members |
+| **C.20** | If you can avoid defining default operations, do (Rule of Zero) |
+| **C.21** | If you define or `=delete` any copy/move/destructor, handle them all (Rule of Five) |
+| **C.35** | Base class destructor: public virtual or protected non-virtual |
+| **C.41** | A constructor should create a fully initialized object |
+| **C.46** | Declare single-argument constructors `explicit` |
+| **C.67** | A polymorphic class should suppress public copy/move |
+| **C.128** | Virtual functions: specify exactly one of `virtual`, `override`, or `final` |
+
+### Rule of Zero
+
+```cpp
+// C.20: Let the compiler generate special members
+struct Employee {
+    std::string name;
+    std::string department;
+    int id;
+    // No destructor, copy/move constructors, or assignment operators needed
+};
+```
+
+### Rule of Five
+
+```cpp
+// C.21: If you must manage a resource, define all five
+class Buffer {
+public:
+    explicit Buffer(std::size_t size)
+        : data_(std::make_unique<char[]>(size)), size_(size) {}
+
+    ~Buffer() = default;
+
+    Buffer(const Buffer& other)
+        : data_(std::make_unique<char[]>(other.size_)), size_(other.size_) {
+        std::copy_n(other.data_.get(), size_, data_.get());
+    }
+
+    Buffer& operator=(const Buffer& other) {
+        if (this != &other) {
+            auto new_data = std::make_unique<char[]>(other.size_);
+            std::copy_n(other.data_.get(), other.size_, new_data.get());
+            data_ = std::move(new_data);
+            size_ = other.size_;
+        }
+        return *this;
+    }
+
+    Buffer(Buffer&&) noexcept = default;
+    Buffer& operator=(Buffer&&) noexcept = default;
+
+private:
+    std::unique_ptr<char[]> data_;
+    std::size_t size_;
+};
+```
+
+### Class Hierarchy
+
+```cpp
+// C.35 + C.128: Virtual destructor, use override
+class Shape {
+public:
+    virtual ~Shape() = default;
+    virtual double area() const = 0;  // C.121: pure interface
+};
+
+class Circle : public Shape {
+public:
+    explicit Circle(double r) : radius_(r) {}
+    double area() const override { return 3.14159 * radius_ * radius_; }
+
+private:
+    double radius_;
+};
+```
+
+### Anti-Patterns
+
+- Calling virtual functions in constructors/destructors (C.82)
+- Using `memset`/`memcpy` on non-trivial types (C.90)
+- Providing different default arguments for virtual function and overrider (C.140)
+- Making data members `const` or references, which suppresses move/copy (C.12)
+
+## Resource Management (R.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **R.1** | Manage resources automatically using RAII |
+| **R.3** | A raw pointer (`T*`) is non-owning |
+| **R.5** | Prefer scoped objects; don't heap-allocate unnecessarily |
+| **R.10** | Avoid `malloc()`/`free()` |
+| **R.11** | Avoid calling `new` and `delete` explicitly |
+| **R.20** | Use `unique_ptr` or `shared_ptr` to represent ownership |
+| **R.21** | Prefer `unique_ptr` over `shared_ptr` unless sharing ownership |
+| **R.22** | Use `make_shared()` to make `shared_ptr`s |
+
+### Smart Pointer Usage
+
+```cpp
+// R.11 + R.20 + R.21: RAII with smart pointers
+auto widget = std::make_unique<Widget>("config");  // unique ownership
+auto cache  = std::make_shared<Cache>(1024);        // shared ownership
+
+// R.3: Raw pointer = non-owning observer
+void render(const Widget* w) {  // does NOT own w
+    if (w) w->draw();
+}
+
+render(widget.get());
+```
+
+### RAII Pattern
+
+```cpp
+// R.1: Resource acquisition is initialization
+class FileHandle {
+public:
+    explicit FileHandle(const std::string& path)
+        : handle_(std::fopen(path.c_str(), "r")) {
+        if (!handle_) throw std::runtime_error("Failed to open: " + path);
+    }
+
+    ~FileHandle() {
+        if (handle_) std::fclose(handle_);
+    }
+
+    FileHandle(const FileHandle&) = delete;
+    FileHandle& operator=(const FileHandle&) = delete;
+    FileHandle(FileHandle&& other) noexcept
+        : handle_(std::exchange(other.handle_, nullptr)) {}
+    FileHandle& operator=(FileHandle&& other) noexcept {
+        if (this != &other) {
+            if (handle_) std::fclose(handle_);
+            handle_ = std::exchange(other.handle_, nullptr);
+        }
+        return *this;
+    }
+
+private:
+    std::FILE* handle_;
+};
+```
+
+### Anti-Patterns
+
+- Naked `new`/`delete` (R.11)
+- `malloc()`/`free()` in C++ code (R.10)
+- Multiple resource allocations in a single expression (R.13 -- exception safety hazard)
+- `shared_ptr` where `unique_ptr` suffices (R.21)
+
+## Expressions & Statements (ES.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **ES.5** | Keep scopes small |
+| **ES.20** | Always initialize an object |
+| **ES.23** | Prefer `{}` initializer syntax |
+| **ES.25** | Declare objects `const` or `constexpr` unless modification is intended |
+| **ES.28** | Use lambdas for complex initialization of `const` variables |
+| **ES.45** | Avoid magic constants; use symbolic constants |
+| **ES.46** | Avoid narrowing/lossy arithmetic conversions |
+| **ES.47** | Use `nullptr` rather than `0` or `NULL` |
+| **ES.48** | Avoid casts |
+| **ES.50** | Don't cast away `const` |
+
+### Initialization
+
+```cpp
+// ES.20 + ES.23 + ES.25: Always initialize, prefer {}, default to const
+const int max_retries{3};
+const std::string name{"widget"};
+const std::vector<int> primes{2, 3, 5, 7, 11};
+
+// ES.28: Lambda for complex const initialization
+const auto config = [&] {
+    Config c;
+    c.timeout = std::chrono::seconds{30};
+    c.retries = max_retries;
+    c.verbose = debug_mode;
+    return c;
+}();
+```
+
+### Anti-Patterns
+
+- Uninitialized variables (ES.20)
+- Using `0` or `NULL` as pointer (ES.47 -- use `nullptr`)
+- C-style casts (ES.48 -- use `static_cast`, `const_cast`, etc.)
+- Casting away `const` (ES.50)
+- Magic numbers without named constants (ES.45)
+- Mixing signed and unsigned arithmetic (ES.100)
+- Reusing names in nested scopes (ES.12)
+
+## Error Handling (E.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **E.1** | Develop an error-handling strategy early in a design |
+| **E.2** | Throw an exception to signal that a function can't perform its assigned task |
+| **E.6** | Use RAII to prevent leaks |
+| **E.12** | Use `noexcept` when throwing is impossible or unacceptable |
+| **E.14** | Use purpose-designed user-defined types as exceptions |
+| **E.15** | Throw by value, catch by reference |
+| **E.16** | Destructors, deallocation, and swap must never fail |
+| **E.17** | Don't try to catch every exception in every function |
+
+### Exception Hierarchy
+
+```cpp
+// E.14 + E.15: Custom exception types, throw by value, catch by reference
+class AppError : public std::runtime_error {
+public:
+    using std::runtime_error::runtime_error;
+};
+
+class NetworkError : public AppError {
+public:
+    NetworkError(const std::string& msg, int code)
+        : AppError(msg), status_code(code) {}
+    int status_code;
+};
+
+void fetch_data(const std::string& url) {
+    // E.2: Throw to signal failure
+    throw NetworkError("connection refused", 503);
+}
+
+void run() {
+    try {
+        fetch_data("https://api.example.com");
+    } catch (const NetworkError& e) {
+        log_error(e.what(), e.status_code);
+    } catch (const AppError& e) {
+        log_error(e.what());
+    }
+    // E.17: Don't catch everything here -- let unexpected errors propagate
+}
+```
+
+### Anti-Patterns
+
+- Throwing built-in types like `int` or string literals (E.14)
+- Catching by value (slicing risk) (E.15)
+- Empty catch blocks that silently swallow errors
+- Using exceptions for flow control (E.3)
+- Error handling based on global state like `errno` (E.28)
+
+## Constants & Immutability (Con.*)
+
+### All Rules
+
+| Rule | Summary |
+|------|---------|
+| **Con.1** | By default, make objects immutable |
+| **Con.2** | By default, make member functions `const` |
+| **Con.3** | By default, pass pointers and references to `const` |
+| **Con.4** | Use `const` for values that don't change after construction |
+| **Con.5** | Use `constexpr` for values computable at compile time |
+
+```cpp
+// Con.1 through Con.5: Immutability by default
+class Sensor {
+public:
+    explicit Sensor(std::string id) : id_(std::move(id)) {}
+
+    // Con.2: const member functions by default
+    const std::string& id() const { return id_; }
+    double last_reading() const { return reading_; }
+
+    // Only non-const when mutation is required
+    void record(double value) { reading_ = value; }
+
+private:
+    const std::string id_;  // Con.4: never changes after construction
+    double reading_{0.0};
+};
+
+// Con.3: Pass by const reference
+void display(const Sensor& s) {
+    std::cout << s.id() << ": " << s.last_reading() << '\n';
+}
+
+// Con.5: Compile-time constants
+constexpr double PI = 3.14159265358979;
+constexpr int MAX_SENSORS = 256;
+```
+
+## Concurrency & Parallelism (CP.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **CP.2** | Avoid data races |
+| **CP.3** | Minimize explicit sharing of writable data |
+| **CP.4** | Think in terms of tasks, rather than threads |
+| **CP.8** | Don't use `volatile` for synchronization |
+| **CP.20** | Use RAII, never plain `lock()`/`unlock()` |
+| **CP.21** | Use `std::scoped_lock` to acquire multiple mutexes |
+| **CP.22** | Never call unknown code while holding a lock |
+| **CP.42** | Don't wait without a condition |
+| **CP.44** | Remember to name your `lock_guard`s and `unique_lock`s |
+| **CP.100** | Don't use lock-free programming unless you absolutely have to |
+
+### Safe Locking
+
+```cpp
+// CP.20 + CP.44: RAII locks, always named
+class ThreadSafeQueue {
+public:
+    void push(int value) {
+        std::lock_guard<std::mutex> lock(mutex_);  // CP.44: named!
+        queue_.push(value);
+        cv_.notify_one();
+    }
+
+    int pop() {
+        std::unique_lock<std::mutex> lock(mutex_);
+        // CP.42: Always wait with a condition
+        cv_.wait(lock, [this] { return !queue_.empty(); });
+        const int value = queue_.front();
+        queue_.pop();
+        return value;
+    }
+
+private:
+    std::mutex mutex_;             // CP.50: mutex with its data
+    std::condition_variable cv_;
+    std::queue<int> queue_;
+};
+```
+
+### Multiple Mutexes
+
+```cpp
+// CP.21: std::scoped_lock for multiple mutexes (deadlock-free)
+void transfer(Account& from, Account& to, double amount) {
+    std::scoped_lock lock(from.mutex_, to.mutex_);
+    from.balance_ -= amount;
+    to.balance_ += amount;
+}
+```
+
+### Anti-Patterns
+
+- `volatile` for synchronization (CP.8 -- it's for hardware I/O only)
+- Detaching threads (CP.26 -- lifetime management becomes nearly impossible)
+- Unnamed lock guards: `std::lock_guard<std::mutex>(m);` destroys immediately (CP.44)
+- Holding locks while calling callbacks (CP.22 -- deadlock risk)
+- Lock-free programming without deep expertise (CP.100)
+
+## Templates & Generic Programming (T.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **T.1** | Use templates to raise the level of abstraction |
+| **T.2** | Use templates to express algorithms for many argument types |
+| **T.10** | Specify concepts for all template arguments |
+| **T.11** | Use standard concepts whenever possible |
+| **T.13** | Prefer shorthand notation for simple concepts |
+| **T.43** | Prefer `using` over `typedef` |
+| **T.120** | Use template metaprogramming only when you really need to |
+| **T.144** | Don't specialize function templates (overload instead) |
+
+### Concepts (C++20)
+
+```cpp
+#include <concepts>
+
+// T.10 + T.11: Constrain templates with standard concepts
+template<std::integral T>
+T gcd(T a, T b) {
+    while (b != 0) {
+        a = std::exchange(b, a % b);
+    }
+    return a;
+}
+
+// T.13: Shorthand concept syntax
+void sort(std::ranges::random_access_range auto& range) {
+    std::ranges::sort(range);
+}
+
+// Custom concept for domain-specific constraints
+template<typename T>
+concept Serializable = requires(const T& t) {
+    { t.serialize() } -> std::convertible_to<std::string>;
+};
+
+template<Serializable T>
+void save(const T& obj, const std::string& path);
+```
+
+### Anti-Patterns
+
+- Unconstrained templates in visible namespaces (T.47)
+- Specializing function templates instead of overloading (T.144)
+- Template metaprogramming where `constexpr` suffices (T.120)
+- `typedef` instead of `using` (T.43)
+
+## Standard Library (SL.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **SL.1** | Use libraries wherever possible |
+| **SL.2** | Prefer the standard library to other libraries |
+| **SL.con.1** | Prefer `std::array` or `std::vector` over C arrays |
+| **SL.con.2** | Prefer `std::vector` by default |
+| **SL.str.1** | Use `std::string` to own character sequences |
+| **SL.str.2** | Use `std::string_view` to refer to character sequences |
+| **SL.io.50** | Avoid `endl` (use `'\n'` -- `endl` forces a flush) |
+
+```cpp
+// SL.con.1 + SL.con.2: Prefer vector/array over C arrays
+const std::array<int, 4> fixed_data{1, 2, 3, 4};
+std::vector<std::string> dynamic_data;
+
+// SL.str.1 + SL.str.2: string owns, string_view observes
+std::string build_greeting(std::string_view name) {
+    return "Hello, " + std::string(name) + "!";
+}
+
+// SL.io.50: Use '\n' not endl
+std::cout << "result: " << value << '\n';
+```
+
+## Enumerations (Enum.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **Enum.1** | Prefer enumerations over macros |
+| **Enum.3** | Prefer `enum class` over plain `enum` |
+| **Enum.5** | Don't use ALL_CAPS for enumerators |
+| **Enum.6** | Avoid unnamed enumerations |
+
+```cpp
+// Enum.3 + Enum.5: Scoped enum, no ALL_CAPS
+enum class Color { red, green, blue };
+enum class LogLevel { debug, info, warning, error };
+
+// BAD: plain enum leaks names, ALL_CAPS clashes with macros
+enum { RED, GREEN, BLUE };           // Enum.3 + Enum.5 + Enum.6 violation
+#define MAX_SIZE 100                  // Enum.1 violation -- use constexpr
+```
+
+## Source Files & Naming (SF.*, NL.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **SF.1** | Use `.cpp` for code files and `.h` for interface files |
+| **SF.7** | Don't write `using namespace` at global scope in a header |
+| **SF.8** | Use `#include` guards for all `.h` files |
+| **SF.11** | Header files should be self-contained |
+| **NL.5** | Avoid encoding type information in names (no Hungarian notation) |
+| **NL.8** | Use a consistent naming style |
+| **NL.9** | Use ALL_CAPS for macro names only |
+| **NL.10** | Prefer `underscore_style` names |
+
+### Header Guard
+
+```cpp
+// SF.8: Include guard (or #pragma once)
+#ifndef PROJECT_MODULE_WIDGET_H
+#define PROJECT_MODULE_WIDGET_H
+
+// SF.11: Self-contained -- include everything this header needs
+#include <string>
+#include <vector>
+
+namespace project::module {
+
+class Widget {
+public:
+    explicit Widget(std::string name);
+    const std::string& name() const;
+
+private:
+    std::string name_;
+};
+
+}  // namespace project::module
+
+#endif  // PROJECT_MODULE_WIDGET_H
+```
+
+### Naming Conventions
+
+```cpp
+// NL.8 + NL.10: Consistent underscore_style
+namespace my_project {
+
+constexpr int max_buffer_size = 4096;  // NL.9: not ALL_CAPS (it's not a macro)
+
+class tcp_connection {                 // underscore_style class
+public:
+    void send_message(std::string_view msg);
+    bool is_connected() const;
+
+private:
+    std::string host_;                 // trailing underscore for members
+    int port_;
+};
+
+}  // namespace my_project
+```
+
+### Anti-Patterns
+
+- `using namespace std;` in a header at global scope (SF.7)
+- Headers that depend on inclusion order (SF.10, SF.11)
+- Hungarian notation like `strName`, `iCount` (NL.5)
+- ALL_CAPS for anything other than macros (NL.9)
+
+## Performance (Per.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **Per.1** | Don't optimize without reason |
+| **Per.2** | Don't optimize prematurely |
+| **Per.6** | Don't make claims about performance without measurements |
+| **Per.7** | Design to enable optimization |
+| **Per.10** | Rely on the static type system |
+| **Per.11** | Move computation from run time to compile time |
+| **Per.19** | Access memory predictably |
+
+### Guidelines
+
+```cpp
+// Per.11: Compile-time computation where possible
+constexpr auto lookup_table = [] {
+    std::array<int, 256> table{};
+    for (int i = 0; i < 256; ++i) {
+        table[i] = i * i;
+    }
+    return table;
+}();
+
+// Per.19: Prefer contiguous data for cache-friendliness
+std::vector<Point> points;           // GOOD: contiguous
+std::vector<std::unique_ptr<Point>> indirect_points; // BAD: pointer chasing
+```
+
+### Anti-Patterns
+
+- Optimizing without profiling data (Per.1, Per.6)
+- Choosing "clever" low-level code over clear abstractions (Per.4, Per.5)
+- Ignoring data layout and cache behavior (Per.19)
+
+## Quick Reference Checklist
+
+Before marking C++ work complete:
+
+- [ ] No raw `new`/`delete` -- use smart pointers or RAII (R.11)
+- [ ] Objects initialized at declaration (ES.20)
+- [ ] Variables are `const`/`constexpr` by default (Con.1, ES.25)
+- [ ] Member functions are `const` where possible (Con.2)
+- [ ] `enum class` instead of plain `enum` (Enum.3)
+- [ ] `nullptr` instead of `0`/`NULL` (ES.47)
+- [ ] No narrowing conversions (ES.46)
+- [ ] No C-style casts (ES.48)
+- [ ] Single-argument constructors are `explicit` (C.46)
+- [ ] Rule of Zero or Rule of Five applied (C.20, C.21)
+- [ ] Base class destructors are public virtual or protected non-virtual (C.35)
+- [ ] Templates are constrained with concepts (T.10)
+- [ ] No `using namespace` in headers at global scope (SF.7)
+- [ ] Headers have include guards and are self-contained (SF.8, SF.11)
+- [ ] Locks use RAII (`scoped_lock`/`lock_guard`) (CP.20)
+- [ ] Exceptions are custom types, thrown by value, caught by reference (E.14, E.15)
+- [ ] `'\n'` instead of `std::endl` (SL.io.50)
+- [ ] No magic numbers (ES.45)

README.md

@@ -222,6 +222,7 @@ everything-claude-code/
 |   |-- verification-loop/          # Continuous verification (Longform Guide)
 |   |-- golang-patterns/            # Go idioms and best practices
 |   |-- golang-testing/             # Go testing patterns, TDD, benchmarks
+|   |-- cpp-coding-standards/         # C++ coding standards from C++ Core Guidelines (NEW)
 |   |-- cpp-testing/                # C++ testing with GoogleTest, CMake/CTest (NEW)
 |   |-- django-patterns/            # Django patterns, models, views (NEW)
 |   |-- django-security/            # Django security best practices (NEW)
@@ -486,6 +487,7 @@ This gives you instant access to all commands, agents, skills, and hooks.
 > git clone https://github.com/affaan-m/everything-claude-code.git
 >
 > # Option A: User-level rules (applies to all projects)
+> mkdir -p ~/.claude/rules
 > cp -r everything-claude-code/rules/common/* ~/.claude/rules/
 > cp -r everything-claude-code/rules/typescript/* ~/.claude/rules/   # pick your stack
 > cp -r everything-claude-code/rules/python/* ~/.claude/rules/

docs/ja-JP/README.md

@@ -454,6 +454,7 @@ Duplicate hooks file detected: ./hooks/hooks.json resolves to already-loaded fil
 > git clone https://github.com/affaan-m/everything-claude-code.git
 >
 > # オプション A：ユーザーレベルルール（すべてのプロジェクトに適用）
+> mkdir -p ~/.claude/rules
 > cp -r everything-claude-code/rules/common/* ~/.claude/rules/
 > cp -r everything-claude-code/rules/typescript/* ~/.claude/rules/   # スタックを選択
 > cp -r everything-claude-code/rules/python/* ~/.claude/rules/

docs/zh-CN/README.md

@@ -456,6 +456,7 @@ Duplicate hooks file detected: ./hooks/hooks.json resolves to already-loaded fil
 > git clone https://github.com/affaan-m/everything-claude-code.git
 >
 > # 选项 A：用户级规则（适用于所有项目）
+> mkdir -p ~/.claude/rules
 > cp -r everything-claude-code/rules/common/* ~/.claude/rules/
 > cp -r everything-claude-code/rules/typescript/* ~/.claude/rules/   # 选择您的技术栈
 > cp -r everything-claude-code/rules/python/* ~/.claude/rules/

hooks/hooks.json

@@ -37,7 +37,7 @@
         "hooks": [
           {
             "type": "command",
-            "command": "node -e \"const fs=require('fs');let d='';process.stdin.on('data',c=>d+=c);process.stdin.on('end',()=>{try{const i=JSON.parse(d);const p=i.tool_input?.file_path||'';if(/\\.(md|txt)$/.test(p)&&!/(README|CLAUDE|AGENTS|CONTRIBUTING)\\.md$/.test(p)){console.error('[Hook] BLOCKED: Unnecessary documentation file creation');console.error('[Hook] File: '+p);console.error('[Hook] Use README.md for documentation instead');process.exit(2)}}catch{}console.log(d)})\""
+            "command": "node -e \"const fs=require('fs');let d='';process.stdin.on('data',c=>d+=c);process.stdin.on('end',()=>{try{const i=JSON.parse(d);const p=i.tool_input?.file_path||'';if(/\\.(md|txt)$/.test(p)&&!/(README|CLAUDE|AGENTS|CONTRIBUTING)\\.md$/.test(p)&&!/\\.claude\\/plans\\//.test(p)){console.error('[Hook] BLOCKED: Unnecessary documentation file creation');console.error('[Hook] File: '+p);console.error('[Hook] Use README.md for documentation instead');process.exit(2)}}catch{}console.log(d)})\""
           }
         ],
         "description": "Block creation of random .md files - keeps docs consolidated"

skills/content-hash-cache-pattern/SKILL.md

@@ -0,0 +1,160 @@
+---
+name: content-hash-cache-pattern
+description: Cache expensive file processing results using SHA-256 content hashes — path-independent, auto-invalidating, with service layer separation.
+---
+
+# Content-Hash File Cache Pattern
+
+Cache expensive file processing results (PDF parsing, text extraction, image analysis) using SHA-256 content hashes as cache keys. Unlike path-based caching, this approach survives file moves/renames and auto-invalidates when content changes.
+
+## When to Activate
+
+- Building file processing pipelines (PDF, images, text extraction)
+- Processing cost is high and same files are processed repeatedly
+- Need a `--cache/--no-cache` CLI option
+- Want to add caching to existing pure functions without modifying them
+
+## Core Pattern
+
+### 1. Content-Hash Based Cache Key
+
+Use file content (not path) as the cache key:
+
+```python
+import hashlib
+from pathlib import Path
+
+_HASH_CHUNK_SIZE = 65536  # 64KB chunks for large files
+
+def compute_file_hash(path: Path) -> str:
+    """SHA-256 of file contents (chunked for large files)."""
+    if not path.is_file():
+        raise FileNotFoundError(f"File not found: {path}")
+    sha256 = hashlib.sha256()
+    with open(path, "rb") as f:
+        while True:
+            chunk = f.read(_HASH_CHUNK_SIZE)
+            if not chunk:
+                break
+            sha256.update(chunk)
+    return sha256.hexdigest()
+```
+
+**Why content hash?** File rename/move = cache hit. Content change = automatic invalidation. No index file needed.
+
+### 2. Frozen Dataclass for Cache Entry
+
+```python
+from dataclasses import dataclass
+
+@dataclass(frozen=True, slots=True)
+class CacheEntry:
+    file_hash: str
+    source_path: str
+    document: ExtractedDocument  # The cached result
+```
+
+### 3. File-Based Cache Storage
+
+Each cache entry is stored as `{hash}.json` — O(1) lookup by hash, no index file required.
+
+```python
+import json
+from typing import Any
+
+def write_cache(cache_dir: Path, entry: CacheEntry) -> None:
+    cache_dir.mkdir(parents=True, exist_ok=True)
+    cache_file = cache_dir / f"{entry.file_hash}.json"
+    data = serialize_entry(entry)
+    cache_file.write_text(json.dumps(data, ensure_ascii=False), encoding="utf-8")
+
+def read_cache(cache_dir: Path, file_hash: str) -> CacheEntry | None:
+    cache_file = cache_dir / f"{file_hash}.json"
+    if not cache_file.is_file():
+        return None
+    try:
+        raw = cache_file.read_text(encoding="utf-8")
+        data = json.loads(raw)
+        return deserialize_entry(data)
+    except (json.JSONDecodeError, ValueError, KeyError):
+        return None  # Treat corruption as cache miss
+```
+
+### 4. Service Layer Wrapper (SRP)
+
+Keep the processing function pure. Add caching as a separate service layer.
+
+```python
+def extract_with_cache(
+    file_path: Path,
+    *,
+    cache_enabled: bool = True,
+    cache_dir: Path = Path(".cache"),
+) -> ExtractedDocument:
+    """Service layer: cache check -> extraction -> cache write."""
+    if not cache_enabled:
+        return extract_text(file_path)  # Pure function, no cache knowledge
+
+    file_hash = compute_file_hash(file_path)
+
+    # Check cache
+    cached = read_cache(cache_dir, file_hash)
+    if cached is not None:
+        logger.info("Cache hit: %s (hash=%s)", file_path.name, file_hash[:12])
+        return cached.document
+
+    # Cache miss -> extract -> store
+    logger.info("Cache miss: %s (hash=%s)", file_path.name, file_hash[:12])
+    doc = extract_text(file_path)
+    entry = CacheEntry(file_hash=file_hash, source_path=str(file_path), document=doc)
+    write_cache(cache_dir, entry)
+    return doc
+```
+
+## Key Design Decisions
+
+| Decision | Rationale |
+|----------|-----------|
+| SHA-256 content hash | Path-independent, auto-invalidates on content change |
+| `{hash}.json` file naming | O(1) lookup, no index file needed |
+| Service layer wrapper | SRP: extraction stays pure, cache is a separate concern |
+| Manual JSON serialization | Full control over frozen dataclass serialization |
+| Corruption returns `None` | Graceful degradation, re-processes on next run |
+| `cache_dir.mkdir(parents=True)` | Lazy directory creation on first write |
+
+## Best Practices
+
+- **Hash content, not paths** — paths change, content identity doesn't
+- **Chunk large files** when hashing — avoid loading entire files into memory
+- **Keep processing functions pure** — they should know nothing about caching
+- **Log cache hit/miss** with truncated hashes for debugging
+- **Handle corruption gracefully** — treat invalid cache entries as misses, never crash
+
+## Anti-Patterns to Avoid
+
+```python
+# BAD: Path-based caching (breaks on file move/rename)
+cache = {"/path/to/file.pdf": result}
+
+# BAD: Adding cache logic inside the processing function (SRP violation)
+def extract_text(path, *, cache_enabled=False, cache_dir=None):
+    if cache_enabled:  # Now this function has two responsibilities
+        ...
+
+# BAD: Using dataclasses.asdict() with nested frozen dataclasses
+# (can cause issues with complex nested types)
+data = dataclasses.asdict(entry)  # Use manual serialization instead
+```
+
+## When to Use
+
+- File processing pipelines (PDF parsing, OCR, text extraction, image analysis)
+- CLI tools that benefit from `--cache/--no-cache` options
+- Batch processing where the same files appear across runs
+- Adding caching to existing pure functions without modifying them
+
+## When NOT to Use
+
+- Data that must always be fresh (real-time feeds)
+- Cache entries that would be extremely large (consider streaming instead)
+- Results that depend on parameters beyond file content (e.g., different extraction configs)

skills/continuous-learning-v2/hooks/observe.sh

@@ -103,7 +103,8 @@ PARSED_OK=$(echo "$PARSED" | python3 -c "import json,sys; print(json.load(sys.st
 if [ "$PARSED_OK" != "True" ]; then
   # Fallback: log raw input for debugging
   timestamp=$(date -u +"%Y-%m-%dT%H:%M:%SZ")
-  TIMESTAMP="$timestamp" echo "$INPUT_JSON" | python3 -c "
+  export TIMESTAMP="$timestamp"
+  echo "$INPUT_JSON" | python3 -c "
 import json, sys, os
 raw = sys.stdin.read()[:2000]
 print(json.dumps({'timestamp': os.environ['TIMESTAMP'], 'event': 'parse_error', 'raw': raw}))
@@ -124,7 +125,8 @@ fi
 # Build and write observation
 timestamp=$(date -u +"%Y-%m-%dT%H:%M:%SZ")
 
-TIMESTAMP="$timestamp" echo "$PARSED" | python3 -c "
+export TIMESTAMP="$timestamp"
+echo "$PARSED" | python3 -c "
 import json, sys, os
 
 parsed = json.load(sys.stdin)

skills/cost-aware-llm-pipeline/SKILL.md

@@ -0,0 +1,182 @@
+---
+name: cost-aware-llm-pipeline
+description: Cost optimization patterns for LLM API usage — model routing by task complexity, budget tracking, retry logic, and prompt caching.
+---
+
+# Cost-Aware LLM Pipeline
+
+Patterns for controlling LLM API costs while maintaining quality. Combines model routing, budget tracking, retry logic, and prompt caching into a composable pipeline.
+
+## When to Activate
+
+- Building applications that call LLM APIs (Claude, GPT, etc.)
+- Processing batches of items with varying complexity
+- Need to stay within a budget for API spend
+- Optimizing cost without sacrificing quality on complex tasks
+
+## Core Concepts
+
+### 1. Model Routing by Task Complexity
+
+Automatically select cheaper models for simple tasks, reserving expensive models for complex ones.
+
+```python
+MODEL_SONNET = "claude-sonnet-4-5-20250929"
+MODEL_HAIKU = "claude-haiku-4-5-20251001"
+
+_SONNET_TEXT_THRESHOLD = 10_000  # chars
+_SONNET_ITEM_THRESHOLD = 30     # items
+
+def select_model(
+    text_length: int,
+    item_count: int,
+    force_model: str | None = None,
+) -> str:
+    """Select model based on task complexity."""
+    if force_model is not None:
+        return force_model
+    if text_length >= _SONNET_TEXT_THRESHOLD or item_count >= _SONNET_ITEM_THRESHOLD:
+        return MODEL_SONNET  # Complex task
+    return MODEL_HAIKU  # Simple task (3-4x cheaper)
+```
+
+### 2. Immutable Cost Tracking
+
+Track cumulative spend with frozen dataclasses. Each API call returns a new tracker — never mutates state.
+
+```python
+from dataclasses import dataclass
+
+@dataclass(frozen=True, slots=True)
+class CostRecord:
+    model: str
+    input_tokens: int
+    output_tokens: int
+    cost_usd: float
+
+@dataclass(frozen=True, slots=True)
+class CostTracker:
+    budget_limit: float = 1.00
+    records: tuple[CostRecord, ...] = ()
+
+    def add(self, record: CostRecord) -> "CostTracker":
+        """Return new tracker with added record (never mutates self)."""
+        return CostTracker(
+            budget_limit=self.budget_limit,
+            records=(*self.records, record),
+        )
+
+    @property
+    def total_cost(self) -> float:
+        return sum(r.cost_usd for r in self.records)
+
+    @property
+    def over_budget(self) -> bool:
+        return self.total_cost > self.budget_limit
+```
+
+### 3. Narrow Retry Logic
+
+Retry only on transient errors. Fail fast on authentication or bad request errors.
+
+```python
+from anthropic import (
+    APIConnectionError,
+    InternalServerError,
+    RateLimitError,
+)
+
+_RETRYABLE_ERRORS = (APIConnectionError, RateLimitError, InternalServerError)
+_MAX_RETRIES = 3
+
+def call_with_retry(func, *, max_retries: int = _MAX_RETRIES):
+    """Retry only on transient errors, fail fast on others."""
+    for attempt in range(max_retries):
+        try:
+            return func()
+        except _RETRYABLE_ERRORS:
+            if attempt == max_retries - 1:
+                raise
+            time.sleep(2 ** attempt)  # Exponential backoff
+    # AuthenticationError, BadRequestError etc. → raise immediately
+```
+
+### 4. Prompt Caching
+
+Cache long system prompts to avoid resending them on every request.
+
+```python
+messages = [
+    {
+        "role": "user",
+        "content": [
+            {
+                "type": "text",
+                "text": system_prompt,
+                "cache_control": {"type": "ephemeral"},  # Cache this
+            },
+            {
+                "type": "text",
+                "text": user_input,  # Variable part
+            },
+        ],
+    }
+]
+```
+
+## Composition
+
+Combine all four techniques in a single pipeline function:
+
+```python
+def process(text: str, config: Config, tracker: CostTracker) -> tuple[Result, CostTracker]:
+    # 1. Route model
+    model = select_model(len(text), estimated_items, config.force_model)
+
+    # 2. Check budget
+    if tracker.over_budget:
+        raise BudgetExceededError(tracker.total_cost, tracker.budget_limit)
+
+    # 3. Call with retry + caching
+    response = call_with_retry(lambda: client.messages.create(
+        model=model,
+        messages=build_cached_messages(system_prompt, text),
+    ))
+
+    # 4. Track cost (immutable)
+    record = CostRecord(model=model, input_tokens=..., output_tokens=..., cost_usd=...)
+    tracker = tracker.add(record)
+
+    return parse_result(response), tracker
+```
+
+## Pricing Reference (2025-2026)
+
+| Model | Input ($/1M tokens) | Output ($/1M tokens) | Relative Cost |
+|-------|---------------------|----------------------|---------------|
+| Haiku 4.5 | $0.80 | $4.00 | 1x |
+| Sonnet 4.5 | $3.00 | $15.00 | ~4x |
+| Opus 4.5 | $15.00 | $75.00 | ~19x |
+
+## Best Practices
+
+- **Start with the cheapest model** and only route to expensive models when complexity thresholds are met
+- **Set explicit budget limits** before processing batches — fail early rather than overspend
+- **Log model selection decisions** so you can tune thresholds based on real data
+- **Use prompt caching** for system prompts over 1024 tokens — saves both cost and latency
+- **Never retry on authentication or validation errors** — only transient failures (network, rate limit, server error)
+
+## Anti-Patterns to Avoid
+
+- Using the most expensive model for all requests regardless of complexity
+- Retrying on all errors (wastes budget on permanent failures)
+- Mutating cost tracking state (makes debugging and auditing difficult)
+- Hardcoding model names throughout the codebase (use constants or config)
+- Ignoring prompt caching for repetitive system prompts
+
+## When to Use
+
+- Any application calling Claude, OpenAI, or similar LLM APIs
+- Batch processing pipelines where cost adds up quickly
+- Multi-model architectures that need intelligent routing
+- Production systems that need budget guardrails

skills/cpp-coding-standards/SKILL.md

@@ -0,0 +1,722 @@
+---
+name: cpp-coding-standards
+description: C++ coding standards based on the C++ Core Guidelines (isocpp.github.io). Use when writing, reviewing, or refactoring C++ code to enforce modern, safe, and idiomatic practices.
+---
+
+# C++ Coding Standards (C++ Core Guidelines)
+
+Comprehensive coding standards for modern C++ (C++17/20/23) derived from the [C++ Core Guidelines](https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines). Enforces type safety, resource safety, immutability, and clarity.
+
+## When to Use
+
+- Writing new C++ code (classes, functions, templates)
+- Reviewing or refactoring existing C++ code
+- Making architectural decisions in C++ projects
+- Enforcing consistent style across a C++ codebase
+- Choosing between language features (e.g., `enum` vs `enum class`, raw pointer vs smart pointer)
+
+### When NOT to Use
+
+- Non-C++ projects
+- Legacy C codebases that cannot adopt modern C++ features
+- Embedded/bare-metal contexts where specific guidelines conflict with hardware constraints (adapt selectively)
+
+## Cross-Cutting Principles
+
+These themes recur across the entire guidelines and form the foundation:
+
+1. **RAII everywhere** (P.8, R.1, E.6, CP.20): Bind resource lifetime to object lifetime
+2. **Immutability by default** (P.10, Con.1-5, ES.25): Start with `const`/`constexpr`; mutability is the exception
+3. **Type safety** (P.4, I.4, ES.46-49, Enum.3): Use the type system to prevent errors at compile time
+4. **Express intent** (P.3, F.1, NL.1-2, T.10): Names, types, and concepts should communicate purpose
+5. **Minimize complexity** (F.2-3, ES.5, Per.4-5): Simple code is correct code
+6. **Value semantics over pointer semantics** (C.10, R.3-5, F.20, CP.31): Prefer returning by value and scoped objects
+
+## Philosophy & Interfaces (P.*, I.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **P.1** | Express ideas directly in code |
+| **P.3** | Express intent |
+| **P.4** | Ideally, a program should be statically type safe |
+| **P.5** | Prefer compile-time checking to run-time checking |
+| **P.8** | Don't leak any resources |
+| **P.10** | Prefer immutable data to mutable data |
+| **I.1** | Make interfaces explicit |
+| **I.2** | Avoid non-const global variables |
+| **I.4** | Make interfaces precisely and strongly typed |
+| **I.11** | Never transfer ownership by a raw pointer or reference |
+| **I.23** | Keep the number of function arguments low |
+
+### DO
+
+```cpp
+// P.10 + I.4: Immutable, strongly typed interface
+struct Temperature {
+    double kelvin;
+};
+
+Temperature boil(const Temperature& water);
+```
+
+### DON'T
+
+```cpp
+// Weak interface: unclear ownership, unclear units
+double boil(double* temp);
+
+// Non-const global variable
+int g_counter = 0;  // I.2 violation
+```
+
+## Functions (F.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **F.1** | Package meaningful operations as carefully named functions |
+| **F.2** | A function should perform a single logical operation |
+| **F.3** | Keep functions short and simple |
+| **F.4** | If a function might be evaluated at compile time, declare it `constexpr` |
+| **F.6** | If your function must not throw, declare it `noexcept` |
+| **F.8** | Prefer pure functions |
+| **F.16** | For "in" parameters, pass cheaply-copied types by value and others by `const&` |
+| **F.20** | For "out" values, prefer return values to output parameters |
+| **F.21** | To return multiple "out" values, prefer returning a struct |
+| **F.43** | Never return a pointer or reference to a local object |
+
+### Parameter Passing
+
+```cpp
+// F.16: Cheap types by value, others by const&
+void print(int x);                           // cheap: by value
+void analyze(const std::string& data);       // expensive: by const&
+void transform(std::string s);               // sink: by value (will move)
+
+// F.20 + F.21: Return values, not output parameters
+struct ParseResult {
+    std::string token;
+    int position;
+};
+
+ParseResult parse(std::string_view input);   // GOOD: return struct
+
+// BAD: output parameters
+void parse(std::string_view input,
+           std::string& token, int& pos);    // avoid this
+```
+
+### Pure Functions and constexpr
+
+```cpp
+// F.4 + F.8: Pure, constexpr where possible
+constexpr int factorial(int n) noexcept {
+    return (n <= 1) ? 1 : n * factorial(n - 1);
+}
+
+static_assert(factorial(5) == 120);
+```
+
+### Anti-Patterns
+
+- Returning `T&&` from functions (F.45)
+- Using `va_arg` / C-style variadics (F.55)
+- Capturing by reference in lambdas passed to other threads (F.53)
+- Returning `const T` which inhibits move semantics (F.49)
+
+## Classes & Class Hierarchies (C.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **C.2** | Use `class` if invariant exists; `struct` if data members vary independently |
+| **C.9** | Minimize exposure of members |
+| **C.20** | If you can avoid defining default operations, do (Rule of Zero) |
+| **C.21** | If you define or `=delete` any copy/move/destructor, handle them all (Rule of Five) |
+| **C.35** | Base class destructor: public virtual or protected non-virtual |
+| **C.41** | A constructor should create a fully initialized object |
+| **C.46** | Declare single-argument constructors `explicit` |
+| **C.67** | A polymorphic class should suppress public copy/move |
+| **C.128** | Virtual functions: specify exactly one of `virtual`, `override`, or `final` |
+
+### Rule of Zero
+
+```cpp
+// C.20: Let the compiler generate special members
+struct Employee {
+    std::string name;
+    std::string department;
+    int id;
+    // No destructor, copy/move constructors, or assignment operators needed
+};
+```
+
+### Rule of Five
+
+```cpp
+// C.21: If you must manage a resource, define all five
+class Buffer {
+public:
+    explicit Buffer(std::size_t size)
+        : data_(std::make_unique<char[]>(size)), size_(size) {}
+
+    ~Buffer() = default;
+
+    Buffer(const Buffer& other)
+        : data_(std::make_unique<char[]>(other.size_)), size_(other.size_) {
+        std::copy_n(other.data_.get(), size_, data_.get());
+    }
+
+    Buffer& operator=(const Buffer& other) {
+        if (this != &other) {
+            auto new_data = std::make_unique<char[]>(other.size_);
+            std::copy_n(other.data_.get(), other.size_, new_data.get());
+            data_ = std::move(new_data);
+            size_ = other.size_;
+        }
+        return *this;
+    }
+
+    Buffer(Buffer&&) noexcept = default;
+    Buffer& operator=(Buffer&&) noexcept = default;
+
+private:
+    std::unique_ptr<char[]> data_;
+    std::size_t size_;
+};
+```
+
+### Class Hierarchy
+
+```cpp
+// C.35 + C.128: Virtual destructor, use override
+class Shape {
+public:
+    virtual ~Shape() = default;
+    virtual double area() const = 0;  // C.121: pure interface
+};
+
+class Circle : public Shape {
+public:
+    explicit Circle(double r) : radius_(r) {}
+    double area() const override { return 3.14159 * radius_ * radius_; }
+
+private:
+    double radius_;
+};
+```
+
+### Anti-Patterns
+
+- Calling virtual functions in constructors/destructors (C.82)
+- Using `memset`/`memcpy` on non-trivial types (C.90)
+- Providing different default arguments for virtual function and overrider (C.140)
+- Making data members `const` or references, which suppresses move/copy (C.12)
+
+## Resource Management (R.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **R.1** | Manage resources automatically using RAII |
+| **R.3** | A raw pointer (`T*`) is non-owning |
+| **R.5** | Prefer scoped objects; don't heap-allocate unnecessarily |
+| **R.10** | Avoid `malloc()`/`free()` |
+| **R.11** | Avoid calling `new` and `delete` explicitly |
+| **R.20** | Use `unique_ptr` or `shared_ptr` to represent ownership |
+| **R.21** | Prefer `unique_ptr` over `shared_ptr` unless sharing ownership |
+| **R.22** | Use `make_shared()` to make `shared_ptr`s |
+
+### Smart Pointer Usage
+
+```cpp
+// R.11 + R.20 + R.21: RAII with smart pointers
+auto widget = std::make_unique<Widget>("config");  // unique ownership
+auto cache  = std::make_shared<Cache>(1024);        // shared ownership
+
+// R.3: Raw pointer = non-owning observer
+void render(const Widget* w) {  // does NOT own w
+    if (w) w->draw();
+}
+
+render(widget.get());
+```
+
+### RAII Pattern
+
+```cpp
+// R.1: Resource acquisition is initialization
+class FileHandle {
+public:
+    explicit FileHandle(const std::string& path)
+        : handle_(std::fopen(path.c_str(), "r")) {
+        if (!handle_) throw std::runtime_error("Failed to open: " + path);
+    }
+
+    ~FileHandle() {
+        if (handle_) std::fclose(handle_);
+    }
+
+    FileHandle(const FileHandle&) = delete;
+    FileHandle& operator=(const FileHandle&) = delete;
+    FileHandle(FileHandle&& other) noexcept
+        : handle_(std::exchange(other.handle_, nullptr)) {}
+    FileHandle& operator=(FileHandle&& other) noexcept {
+        if (this != &other) {
+            if (handle_) std::fclose(handle_);
+            handle_ = std::exchange(other.handle_, nullptr);
+        }
+        return *this;
+    }
+
+private:
+    std::FILE* handle_;
+};
+```
+
+### Anti-Patterns
+
+- Naked `new`/`delete` (R.11)
+- `malloc()`/`free()` in C++ code (R.10)
+- Multiple resource allocations in a single expression (R.13 -- exception safety hazard)
+- `shared_ptr` where `unique_ptr` suffices (R.21)
+
+## Expressions & Statements (ES.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **ES.5** | Keep scopes small |
+| **ES.20** | Always initialize an object |
+| **ES.23** | Prefer `{}` initializer syntax |
+| **ES.25** | Declare objects `const` or `constexpr` unless modification is intended |
+| **ES.28** | Use lambdas for complex initialization of `const` variables |
+| **ES.45** | Avoid magic constants; use symbolic constants |
+| **ES.46** | Avoid narrowing/lossy arithmetic conversions |
+| **ES.47** | Use `nullptr` rather than `0` or `NULL` |
+| **ES.48** | Avoid casts |
+| **ES.50** | Don't cast away `const` |
+
+### Initialization
+
+```cpp
+// ES.20 + ES.23 + ES.25: Always initialize, prefer {}, default to const
+const int max_retries{3};
+const std::string name{"widget"};
+const std::vector<int> primes{2, 3, 5, 7, 11};
+
+// ES.28: Lambda for complex const initialization
+const auto config = [&] {
+    Config c;
+    c.timeout = std::chrono::seconds{30};
+    c.retries = max_retries;
+    c.verbose = debug_mode;
+    return c;
+}();
+```
+
+### Anti-Patterns
+
+- Uninitialized variables (ES.20)
+- Using `0` or `NULL` as pointer (ES.47 -- use `nullptr`)
+- C-style casts (ES.48 -- use `static_cast`, `const_cast`, etc.)
+- Casting away `const` (ES.50)
+- Magic numbers without named constants (ES.45)
+- Mixing signed and unsigned arithmetic (ES.100)
+- Reusing names in nested scopes (ES.12)
+
+## Error Handling (E.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **E.1** | Develop an error-handling strategy early in a design |
+| **E.2** | Throw an exception to signal that a function can't perform its assigned task |
+| **E.6** | Use RAII to prevent leaks |
+| **E.12** | Use `noexcept` when throwing is impossible or unacceptable |
+| **E.14** | Use purpose-designed user-defined types as exceptions |
+| **E.15** | Throw by value, catch by reference |
+| **E.16** | Destructors, deallocation, and swap must never fail |
+| **E.17** | Don't try to catch every exception in every function |
+
+### Exception Hierarchy
+
+```cpp
+// E.14 + E.15: Custom exception types, throw by value, catch by reference
+class AppError : public std::runtime_error {
+public:
+    using std::runtime_error::runtime_error;
+};
+
+class NetworkError : public AppError {
+public:
+    NetworkError(const std::string& msg, int code)
+        : AppError(msg), status_code(code) {}
+    int status_code;
+};
+
+void fetch_data(const std::string& url) {
+    // E.2: Throw to signal failure
+    throw NetworkError("connection refused", 503);
+}
+
+void run() {
+    try {
+        fetch_data("https://api.example.com");
+    } catch (const NetworkError& e) {
+        log_error(e.what(), e.status_code);
+    } catch (const AppError& e) {
+        log_error(e.what());
+    }
+    // E.17: Don't catch everything here -- let unexpected errors propagate
+}
+```
+
+### Anti-Patterns
+
+- Throwing built-in types like `int` or string literals (E.14)
+- Catching by value (slicing risk) (E.15)
+- Empty catch blocks that silently swallow errors
+- Using exceptions for flow control (E.3)
+- Error handling based on global state like `errno` (E.28)
+
+## Constants & Immutability (Con.*)
+
+### All Rules
+
+| Rule | Summary |
+|------|---------|
+| **Con.1** | By default, make objects immutable |
+| **Con.2** | By default, make member functions `const` |
+| **Con.3** | By default, pass pointers and references to `const` |
+| **Con.4** | Use `const` for values that don't change after construction |
+| **Con.5** | Use `constexpr` for values computable at compile time |
+
+```cpp
+// Con.1 through Con.5: Immutability by default
+class Sensor {
+public:
+    explicit Sensor(std::string id) : id_(std::move(id)) {}
+
+    // Con.2: const member functions by default
+    const std::string& id() const { return id_; }
+    double last_reading() const { return reading_; }
+
+    // Only non-const when mutation is required
+    void record(double value) { reading_ = value; }
+
+private:
+    const std::string id_;  // Con.4: never changes after construction
+    double reading_{0.0};
+};
+
+// Con.3: Pass by const reference
+void display(const Sensor& s) {
+    std::cout << s.id() << ": " << s.last_reading() << '\n';
+}
+
+// Con.5: Compile-time constants
+constexpr double PI = 3.14159265358979;
+constexpr int MAX_SENSORS = 256;
+```
+
+## Concurrency & Parallelism (CP.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **CP.2** | Avoid data races |
+| **CP.3** | Minimize explicit sharing of writable data |
+| **CP.4** | Think in terms of tasks, rather than threads |
+| **CP.8** | Don't use `volatile` for synchronization |
+| **CP.20** | Use RAII, never plain `lock()`/`unlock()` |
+| **CP.21** | Use `std::scoped_lock` to acquire multiple mutexes |
+| **CP.22** | Never call unknown code while holding a lock |
+| **CP.42** | Don't wait without a condition |
+| **CP.44** | Remember to name your `lock_guard`s and `unique_lock`s |
+| **CP.100** | Don't use lock-free programming unless you absolutely have to |
+
+### Safe Locking
+
+```cpp
+// CP.20 + CP.44: RAII locks, always named
+class ThreadSafeQueue {
+public:
+    void push(int value) {
+        std::lock_guard<std::mutex> lock(mutex_);  // CP.44: named!
+        queue_.push(value);
+        cv_.notify_one();
+    }
+
+    int pop() {
+        std::unique_lock<std::mutex> lock(mutex_);
+        // CP.42: Always wait with a condition
+        cv_.wait(lock, [this] { return !queue_.empty(); });
+        const int value = queue_.front();
+        queue_.pop();
+        return value;
+    }
+
+private:
+    std::mutex mutex_;             // CP.50: mutex with its data
+    std::condition_variable cv_;
+    std::queue<int> queue_;
+};
+```
+
+### Multiple Mutexes
+
+```cpp
+// CP.21: std::scoped_lock for multiple mutexes (deadlock-free)
+void transfer(Account& from, Account& to, double amount) {
+    std::scoped_lock lock(from.mutex_, to.mutex_);
+    from.balance_ -= amount;
+    to.balance_ += amount;
+}
+```
+
+### Anti-Patterns
+
+- `volatile` for synchronization (CP.8 -- it's for hardware I/O only)
+- Detaching threads (CP.26 -- lifetime management becomes nearly impossible)
+- Unnamed lock guards: `std::lock_guard<std::mutex>(m);` destroys immediately (CP.44)
+- Holding locks while calling callbacks (CP.22 -- deadlock risk)
+- Lock-free programming without deep expertise (CP.100)
+
+## Templates & Generic Programming (T.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **T.1** | Use templates to raise the level of abstraction |
+| **T.2** | Use templates to express algorithms for many argument types |
+| **T.10** | Specify concepts for all template arguments |
+| **T.11** | Use standard concepts whenever possible |
+| **T.13** | Prefer shorthand notation for simple concepts |
+| **T.43** | Prefer `using` over `typedef` |
+| **T.120** | Use template metaprogramming only when you really need to |
+| **T.144** | Don't specialize function templates (overload instead) |
+
+### Concepts (C++20)
+
+```cpp
+#include <concepts>
+
+// T.10 + T.11: Constrain templates with standard concepts
+template<std::integral T>
+T gcd(T a, T b) {
+    while (b != 0) {
+        a = std::exchange(b, a % b);
+    }
+    return a;
+}
+
+// T.13: Shorthand concept syntax
+void sort(std::ranges::random_access_range auto& range) {
+    std::ranges::sort(range);
+}
+
+// Custom concept for domain-specific constraints
+template<typename T>
+concept Serializable = requires(const T& t) {
+    { t.serialize() } -> std::convertible_to<std::string>;
+};
+
+template<Serializable T>
+void save(const T& obj, const std::string& path);
+```
+
+### Anti-Patterns
+
+- Unconstrained templates in visible namespaces (T.47)
+- Specializing function templates instead of overloading (T.144)
+- Template metaprogramming where `constexpr` suffices (T.120)
+- `typedef` instead of `using` (T.43)
+
+## Standard Library (SL.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **SL.1** | Use libraries wherever possible |
+| **SL.2** | Prefer the standard library to other libraries |
+| **SL.con.1** | Prefer `std::array` or `std::vector` over C arrays |
+| **SL.con.2** | Prefer `std::vector` by default |
+| **SL.str.1** | Use `std::string` to own character sequences |
+| **SL.str.2** | Use `std::string_view` to refer to character sequences |
+| **SL.io.50** | Avoid `endl` (use `'\n'` -- `endl` forces a flush) |
+
+```cpp
+// SL.con.1 + SL.con.2: Prefer vector/array over C arrays
+const std::array<int, 4> fixed_data{1, 2, 3, 4};
+std::vector<std::string> dynamic_data;
+
+// SL.str.1 + SL.str.2: string owns, string_view observes
+std::string build_greeting(std::string_view name) {
+    return "Hello, " + std::string(name) + "!";
+}
+
+// SL.io.50: Use '\n' not endl
+std::cout << "result: " << value << '\n';
+```
+
+## Enumerations (Enum.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **Enum.1** | Prefer enumerations over macros |
+| **Enum.3** | Prefer `enum class` over plain `enum` |
+| **Enum.5** | Don't use ALL_CAPS for enumerators |
+| **Enum.6** | Avoid unnamed enumerations |
+
+```cpp
+// Enum.3 + Enum.5: Scoped enum, no ALL_CAPS
+enum class Color { red, green, blue };
+enum class LogLevel { debug, info, warning, error };
+
+// BAD: plain enum leaks names, ALL_CAPS clashes with macros
+enum { RED, GREEN, BLUE };           // Enum.3 + Enum.5 + Enum.6 violation
+#define MAX_SIZE 100                  // Enum.1 violation -- use constexpr
+```
+
+## Source Files & Naming (SF.*, NL.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **SF.1** | Use `.cpp` for code files and `.h` for interface files |
+| **SF.7** | Don't write `using namespace` at global scope in a header |
+| **SF.8** | Use `#include` guards for all `.h` files |
+| **SF.11** | Header files should be self-contained |
+| **NL.5** | Avoid encoding type information in names (no Hungarian notation) |
+| **NL.8** | Use a consistent naming style |
+| **NL.9** | Use ALL_CAPS for macro names only |
+| **NL.10** | Prefer `underscore_style` names |
+
+### Header Guard
+
+```cpp
+// SF.8: Include guard (or #pragma once)
+#ifndef PROJECT_MODULE_WIDGET_H
+#define PROJECT_MODULE_WIDGET_H
+
+// SF.11: Self-contained -- include everything this header needs
+#include <string>
+#include <vector>
+
+namespace project::module {
+
+class Widget {
+public:
+    explicit Widget(std::string name);
+    const std::string& name() const;
+
+private:
+    std::string name_;
+};
+
+}  // namespace project::module
+
+#endif  // PROJECT_MODULE_WIDGET_H
+```
+
+### Naming Conventions
+
+```cpp
+// NL.8 + NL.10: Consistent underscore_style
+namespace my_project {
+
+constexpr int max_buffer_size = 4096;  // NL.9: not ALL_CAPS (it's not a macro)
+
+class tcp_connection {                 // underscore_style class
+public:
+    void send_message(std::string_view msg);
+    bool is_connected() const;
+
+private:
+    std::string host_;                 // trailing underscore for members
+    int port_;
+};
+
+}  // namespace my_project
+```
+
+### Anti-Patterns
+
+- `using namespace std;` in a header at global scope (SF.7)
+- Headers that depend on inclusion order (SF.10, SF.11)
+- Hungarian notation like `strName`, `iCount` (NL.5)
+- ALL_CAPS for anything other than macros (NL.9)
+
+## Performance (Per.*)
+
+### Key Rules
+
+| Rule | Summary |
+|------|---------|
+| **Per.1** | Don't optimize without reason |
+| **Per.2** | Don't optimize prematurely |
+| **Per.6** | Don't make claims about performance without measurements |
+| **Per.7** | Design to enable optimization |
+| **Per.10** | Rely on the static type system |
+| **Per.11** | Move computation from run time to compile time |
+| **Per.19** | Access memory predictably |
+
+### Guidelines
+
+```cpp
+// Per.11: Compile-time computation where possible
+constexpr auto lookup_table = [] {
+    std::array<int, 256> table{};
+    for (int i = 0; i < 256; ++i) {
+        table[i] = i * i;
+    }
+    return table;
+}();
+
+// Per.19: Prefer contiguous data for cache-friendliness
+std::vector<Point> points;           // GOOD: contiguous
+std::vector<std::unique_ptr<Point>> indirect_points; // BAD: pointer chasing
+```
+
+### Anti-Patterns
+
+- Optimizing without profiling data (Per.1, Per.6)
+- Choosing "clever" low-level code over clear abstractions (Per.4, Per.5)
+- Ignoring data layout and cache behavior (Per.19)
+
+## Quick Reference Checklist
+
+Before marking C++ work complete:
+
+- [ ] No raw `new`/`delete` -- use smart pointers or RAII (R.11)
+- [ ] Objects initialized at declaration (ES.20)
+- [ ] Variables are `const`/`constexpr` by default (Con.1, ES.25)
+- [ ] Member functions are `const` where possible (Con.2)
+- [ ] `enum class` instead of plain `enum` (Enum.3)
+- [ ] `nullptr` instead of `0`/`NULL` (ES.47)
+- [ ] No narrowing conversions (ES.46)
+- [ ] No C-style casts (ES.48)
+- [ ] Single-argument constructors are `explicit` (C.46)
+- [ ] Rule of Zero or Rule of Five applied (C.20, C.21)
+- [ ] Base class destructors are public virtual or protected non-virtual (C.35)
+- [ ] Templates are constrained with concepts (T.10)
+- [ ] No `using namespace` in headers at global scope (SF.7)
+- [ ] Headers have include guards and are self-contained (SF.8, SF.11)
+- [ ] Locks use RAII (`scoped_lock`/`lock_guard`) (CP.20)
+- [ ] Exceptions are custom types, thrown by value, caught by reference (E.14, E.15)
+- [ ] `'\n'` instead of `std::endl` (SL.io.50)
+- [ ] No magic numbers (ES.45)

skills/regex-vs-llm-structured-text/SKILL.md

@@ -0,0 +1,219 @@
+---
+name: regex-vs-llm-structured-text
+description: Decision framework for choosing between regex and LLM when parsing structured text — start with regex, add LLM only for low-confidence edge cases.
+---
+
+# Regex vs LLM for Structured Text Parsing
+
+A practical decision framework for parsing structured text (quizzes, forms, invoices, documents). The key insight: regex handles 95-98% of cases cheaply and deterministically. Reserve expensive LLM calls for the remaining edge cases.
+
+## When to Activate
+
+- Parsing structured text with repeating patterns (questions, forms, tables)
+- Deciding between regex and LLM for text extraction
+- Building hybrid pipelines that combine both approaches
+- Optimizing cost/accuracy tradeoffs in text processing
+
+## Decision Framework
+
+```
+Is the text format consistent and repeating?
+├── Yes (>90% follows a pattern) → Start with Regex
+│   ├── Regex handles 95%+ → Done, no LLM needed
+│   └── Regex handles <95% → Add LLM for edge cases only
+└── No (free-form, highly variable) → Use LLM directly
+```
+
+## Architecture Pattern
+
+```
+Source Text
+    │
+    ▼
+[Regex Parser] ─── Extracts structure (95-98% accuracy)
+    │
+    ▼
+[Text Cleaner] ─── Removes noise (markers, page numbers, artifacts)
+    │
+    ▼
+[Confidence Scorer] ─── Flags low-confidence extractions
+    │
+    ├── High confidence (≥0.95) → Direct output
+    │
+    └── Low confidence (<0.95) → [LLM Validator] → Output
+```
+
+## Implementation
+
+### 1. Regex Parser (Handles the Majority)
+
+```python
+import re
+from dataclasses import dataclass
+
+@dataclass(frozen=True)
+class ParsedItem:
+    id: str
+    text: str
+    choices: tuple[str, ...]
+    answer: str
+    confidence: float = 1.0
+
+def parse_structured_text(content: str) -> list[ParsedItem]:
+    """Parse structured text using regex patterns."""
+    pattern = re.compile(
+        r"(?P<id>\d+)\.\s*(?P<text>.+?)\n"
+        r"(?P<choices>(?:[A-D]\..+?\n)+)"
+        r"Answer:\s*(?P<answer>[A-D])",
+        re.MULTILINE | re.DOTALL,
+    )
+    items = []
+    for match in pattern.finditer(content):
+        choices = tuple(
+            c.strip() for c in re.findall(r"[A-D]\.\s*(.+)", match.group("choices"))
+        )
+        items.append(ParsedItem(
+            id=match.group("id"),
+            text=match.group("text").strip(),
+            choices=choices,
+            answer=match.group("answer"),
+        ))
+    return items
+```
+
+### 2. Confidence Scoring
+
+Flag items that may need LLM review:
+
+```python
+@dataclass(frozen=True)
+class ConfidenceFlag:
+    item_id: str
+    score: float
+    reasons: tuple[str, ...]
+
+def score_confidence(item: ParsedItem) -> ConfidenceFlag:
+    """Score extraction confidence and flag issues."""
+    reasons = []
+    score = 1.0
+
+    if len(item.choices) < 3:
+        reasons.append("few_choices")
+        score -= 0.3
+
+    if not item.answer:
+        reasons.append("missing_answer")
+        score -= 0.5
+
+    if len(item.text) < 10:
+        reasons.append("short_text")
+        score -= 0.2
+
+    return ConfidenceFlag(
+        item_id=item.id,
+        score=max(0.0, score),
+        reasons=tuple(reasons),
+    )
+
+def identify_low_confidence(
+    items: list[ParsedItem],
+    threshold: float = 0.95,
+) -> list[ConfidenceFlag]:
+    """Return items below confidence threshold."""
+    flags = [score_confidence(item) for item in items]
+    return [f for f in flags if f.score < threshold]
+```
+
+### 3. LLM Validator (Edge Cases Only)
+
+```python
+def validate_with_llm(
+    item: ParsedItem,
+    original_text: str,
+    client,
+) -> ParsedItem:
+    """Use LLM to fix low-confidence extractions."""
+    response = client.messages.create(
+        model="claude-haiku-4-5-20251001",  # Cheapest model for validation
+        max_tokens=500,
+        messages=[{
+            "role": "user",
+            "content": (
+                f"Extract the question, choices, and answer from this text.\n\n"
+                f"Text: {original_text}\n\n"
+                f"Current extraction: {item}\n\n"
+                f"Return corrected JSON if needed, or 'CORRECT' if accurate."
+            ),
+        }],
+    )
+    # Parse LLM response and return corrected item...
+    return corrected_item
+```
+
+### 4. Hybrid Pipeline
+
+```python
+def process_document(
+    content: str,
+    *,
+    llm_client=None,
+    confidence_threshold: float = 0.95,
+) -> list[ParsedItem]:
+    """Full pipeline: regex -> confidence check -> LLM for edge cases."""
+    # Step 1: Regex extraction (handles 95-98%)
+    items = parse_structured_text(content)
+
+    # Step 2: Confidence scoring
+    low_confidence = identify_low_confidence(items, confidence_threshold)
+
+    if not low_confidence or llm_client is None:
+        return items
+
+    # Step 3: LLM validation (only for flagged items)
+    low_conf_ids = {f.item_id for f in low_confidence}
+    result = []
+    for item in items:
+        if item.id in low_conf_ids:
+            result.append(validate_with_llm(item, content, llm_client))
+        else:
+            result.append(item)
+
+    return result
+```
+
+## Real-World Metrics
+
+From a production quiz parsing pipeline (410 items):
+
+| Metric | Value |
+|--------|-------|
+| Regex success rate | 98.0% |
+| Low confidence items | 8 (2.0%) |
+| LLM calls needed | ~5 |
+| Cost savings vs all-LLM | ~95% |
+| Test coverage | 93% |
+
+## Best Practices
+
+- **Start with regex** — even imperfect regex gives you a baseline to improve
+- **Use confidence scoring** to programmatically identify what needs LLM help
+- **Use the cheapest LLM** for validation (Haiku-class models are sufficient)
+- **Never mutate** parsed items — return new instances from cleaning/validation steps
+- **TDD works well** for parsers — write tests for known patterns first, then edge cases
+- **Log metrics** (regex success rate, LLM call count) to track pipeline health
+
+## Anti-Patterns to Avoid
+
+- Sending all text to an LLM when regex handles 95%+ of cases (expensive and slow)
+- Using regex for free-form, highly variable text (LLM is better here)
+- Skipping confidence scoring and hoping regex "just works"
+- Mutating parsed objects during cleaning/validation steps
+- Not testing edge cases (malformed input, missing fields, encoding issues)
+
+## When to Use
+
+- Quiz/exam question parsing
+- Form data extraction
+- Invoice/receipt processing
+- Document structure parsing (headers, sections, tables)
+- Any structured text with repeating patterns where cost matters

skills/swift-actor-persistence/SKILL.md

@@ -0,0 +1,142 @@
+---
+name: swift-actor-persistence
+description: Thread-safe data persistence in Swift using actors — in-memory cache with file-backed storage, eliminating data races by design.
+---
+
+# Swift Actors for Thread-Safe Persistence
+
+Patterns for building thread-safe data persistence layers using Swift actors. Combines in-memory caching with file-backed storage, leveraging the actor model to eliminate data races at compile time.
+
+## When to Activate
+
+- Building a data persistence layer in Swift 5.5+
+- Need thread-safe access to shared mutable state
+- Want to eliminate manual synchronization (locks, DispatchQueues)
+- Building offline-first apps with local storage
+
+## Core Pattern
+
+### Actor-Based Repository
+
+The actor model guarantees serialized access — no data races, enforced by the compiler.
+
+```swift
+public actor LocalRepository<T: Codable & Identifiable> where T.ID == String {
+    private var cache: [String: T] = [:]
+    private let fileURL: URL
+
+    public init(directory: URL = .documentsDirectory, filename: String = "data.json") {
+        self.fileURL = directory.appendingPathComponent(filename)
+        // Synchronous load during init (actor isolation not yet active)
+        self.cache = Self.loadSynchronously(from: fileURL)
+    }
+
+    // MARK: - Public API
+
+    public func save(_ item: T) throws {
+        cache[item.id] = item
+        try persistToFile()
+    }
+
+    public func delete(_ id: String) throws {
+        cache[id] = nil
+        try persistToFile()
+    }
+
+    public func find(by id: String) -> T? {
+        cache[id]
+    }
+
+    public func loadAll() -> [T] {
+        Array(cache.values)
+    }
+
+    // MARK: - Private
+
+    private func persistToFile() throws {
+        let data = try JSONEncoder().encode(Array(cache.values))
+        try data.write(to: fileURL, options: .atomic)
+    }
+
+    private static func loadSynchronously(from url: URL) -> [String: T] {
+        guard let data = try? Data(contentsOf: url),
+              let items = try? JSONDecoder().decode([T].self, from: data) else {
+            return [:]
+        }
+        return Dictionary(uniqueKeysWithValues: items.map { ($0.id, $0) })
+    }
+}
+```
+
+### Usage
+
+All calls are automatically async due to actor isolation:
+
+```swift
+let repository = LocalRepository<Question>()
+
+// Read — fast O(1) lookup from in-memory cache
+let question = await repository.find(by: "q-001")
+let allQuestions = await repository.loadAll()
+
+// Write — updates cache and persists to file atomically
+try await repository.save(newQuestion)
+try await repository.delete("q-001")
+```
+
+### Combining with @Observable ViewModel
+
+```swift
+@Observable
+final class QuestionListViewModel {
+    private(set) var questions: [Question] = []
+    private let repository: LocalRepository<Question>
+
+    init(repository: LocalRepository<Question> = LocalRepository()) {
+        self.repository = repository
+    }
+
+    func load() async {
+        questions = await repository.loadAll()
+    }
+
+    func add(_ question: Question) async throws {
+        try await repository.save(question)
+        questions = await repository.loadAll()
+    }
+}
+```
+
+## Key Design Decisions
+
+| Decision | Rationale |
+|----------|-----------|
+| Actor (not class + lock) | Compiler-enforced thread safety, no manual synchronization |
+| In-memory cache + file persistence | Fast reads from cache, durable writes to disk |
+| Synchronous init loading | Avoids async initialization complexity |
+| Dictionary keyed by ID | O(1) lookups by identifier |
+| Generic over `Codable & Identifiable` | Reusable across any model type |
+| Atomic file writes (`.atomic`) | Prevents partial writes on crash |
+
+## Best Practices
+
+- **Use `Sendable` types** for all data crossing actor boundaries
+- **Keep the actor's public API minimal** — only expose domain operations, not persistence details
+- **Use `.atomic` writes** to prevent data corruption if the app crashes mid-write
+- **Load synchronously in `init`** — async initializers add complexity with minimal benefit for local files
+- **Combine with `@Observable`** ViewModels for reactive UI updates
+
+## Anti-Patterns to Avoid
+
+- Using `DispatchQueue` or `NSLock` instead of actors for new Swift concurrency code
+- Exposing the internal cache dictionary to external callers
+- Making the file URL configurable without validation
+- Forgetting that all actor method calls are `await` — callers must handle async context
+- Using `nonisolated` to bypass actor isolation (defeats the purpose)
+
+## When to Use
+
+- Local data storage in iOS/macOS apps (user data, settings, cached content)
+- Offline-first architectures that sync to a server later
+- Any shared mutable state that multiple parts of the app access concurrently
+- Replacing legacy `DispatchQueue`-based thread safety with modern Swift concurrency

skills/swift-protocol-di-testing/SKILL.md

@@ -0,0 +1,189 @@
+---
+name: swift-protocol-di-testing
+description: Protocol-based dependency injection for testable Swift code — mock file system, network, and external APIs using focused protocols and Swift Testing.
+---
+
+# Swift Protocol-Based Dependency Injection for Testing
+
+Patterns for making Swift code testable by abstracting external dependencies (file system, network, iCloud) behind small, focused protocols. Enables deterministic tests without I/O.
+
+## When to Activate
+
+- Writing Swift code that accesses file system, network, or external APIs
+- Need to test error handling paths without triggering real failures
+- Building modules that work across environments (app, test, SwiftUI preview)
+- Designing testable architecture with Swift concurrency (actors, Sendable)
+
+## Core Pattern
+
+### 1. Define Small, Focused Protocols
+
+Each protocol handles exactly one external concern.
+
+```swift
+// File system access
+public protocol FileSystemProviding: Sendable {
+    func containerURL(for purpose: Purpose) -> URL?
+}
+
+// File read/write operations
+public protocol FileAccessorProviding: Sendable {
+    func read(from url: URL) throws -> Data
+    func write(_ data: Data, to url: URL) throws
+    func fileExists(at url: URL) -> Bool
+}
+
+// Bookmark storage (e.g., for sandboxed apps)
+public protocol BookmarkStorageProviding: Sendable {
+    func saveBookmark(_ data: Data, for key: String) throws
+    func loadBookmark(for key: String) throws -> Data?
+}
+```
+
+### 2. Create Default (Production) Implementations
+
+```swift
+public struct DefaultFileSystemProvider: FileSystemProviding {
+    public init() {}
+
+    public func containerURL(for purpose: Purpose) -> URL? {
+        FileManager.default.url(forUbiquityContainerIdentifier: nil)
+    }
+}
+
+public struct DefaultFileAccessor: FileAccessorProviding {
+    public init() {}
+
+    public func read(from url: URL) throws -> Data {
+        try Data(contentsOf: url)
+    }
+
+    public func write(_ data: Data, to url: URL) throws {
+        try data.write(to: url, options: .atomic)
+    }
+
+    public func fileExists(at url: URL) -> Bool {
+        FileManager.default.fileExists(atPath: url.path)
+    }
+}
+```
+
+### 3. Create Mock Implementations for Testing
+
+```swift
+public final class MockFileAccessor: FileAccessorProviding, @unchecked Sendable {
+    public var files: [URL: Data] = [:]
+    public var readError: Error?
+    public var writeError: Error?
+
+    public init() {}
+
+    public func read(from url: URL) throws -> Data {
+        if let error = readError { throw error }
+        guard let data = files[url] else {
+            throw CocoaError(.fileReadNoSuchFile)
+        }
+        return data
+    }
+
+    public func write(_ data: Data, to url: URL) throws {
+        if let error = writeError { throw error }
+        files[url] = data
+    }
+
+    public func fileExists(at url: URL) -> Bool {
+        files[url] != nil
+    }
+}
+```
+
+### 4. Inject Dependencies with Default Parameters
+
+Production code uses defaults; tests inject mocks.
+
+```swift
+public actor SyncManager {
+    private let fileSystem: FileSystemProviding
+    private let fileAccessor: FileAccessorProviding
+
+    public init(
+        fileSystem: FileSystemProviding = DefaultFileSystemProvider(),
+        fileAccessor: FileAccessorProviding = DefaultFileAccessor()
+    ) {
+        self.fileSystem = fileSystem
+        self.fileAccessor = fileAccessor
+    }
+
+    public func sync() async throws {
+        guard let containerURL = fileSystem.containerURL(for: .sync) else {
+            throw SyncError.containerNotAvailable
+        }
+        let data = try fileAccessor.read(
+            from: containerURL.appendingPathComponent("data.json")
+        )
+        // Process data...
+    }
+}
+```
+
+### 5. Write Tests with Swift Testing
+
+```swift
+import Testing
+
+@Test("Sync manager handles missing container")
+func testMissingContainer() async {
+    let mockFileSystem = MockFileSystemProvider(containerURL: nil)
+    let manager = SyncManager(fileSystem: mockFileSystem)
+
+    await #expect(throws: SyncError.containerNotAvailable) {
+        try await manager.sync()
+    }
+}
+
+@Test("Sync manager reads data correctly")
+func testReadData() async throws {
+    let mockFileAccessor = MockFileAccessor()
+    mockFileAccessor.files[testURL] = testData
+
+    let manager = SyncManager(fileAccessor: mockFileAccessor)
+    let result = try await manager.loadData()
+
+    #expect(result == expectedData)
+}
+
+@Test("Sync manager handles read errors gracefully")
+func testReadError() async {
+    let mockFileAccessor = MockFileAccessor()
+    mockFileAccessor.readError = CocoaError(.fileReadCorruptFile)
+
+    let manager = SyncManager(fileAccessor: mockFileAccessor)
+
+    await #expect(throws: SyncError.self) {
+        try await manager.sync()
+    }
+}
+```
+
+## Best Practices
+
+- **Single Responsibility**: Each protocol should handle one concern — don't create "god protocols" with many methods
+- **Sendable conformance**: Required when protocols are used across actor boundaries
+- **Default parameters**: Let production code use real implementations by default; only tests need to specify mocks
+- **Error simulation**: Design mocks with configurable error properties for testing failure paths
+- **Only mock boundaries**: Mock external dependencies (file system, network, APIs), not internal types
+
+## Anti-Patterns to Avoid
+
+- Creating a single large protocol that covers all external access
+- Mocking internal types that have no external dependencies
+- Using `#if DEBUG` conditionals instead of proper dependency injection
+- Forgetting `Sendable` conformance when used with actors
+- Over-engineering: if a type has no external dependencies, it doesn't need a protocol
+
+## When to Use
+
+- Any Swift code that touches file system, network, or external APIs
+- Testing error handling paths that are hard to trigger in real environments
+- Building modules that need to work in app, test, and SwiftUI preview contexts
+- Apps using Swift concurrency (actors, structured concurrency) that need testable architecture