icore 1.0.19 → 2.0.0

package/readme.md

@@ -5,713 +5,243 @@
 [![types](https://img.shields.io/npm/types/icore.svg)](https://www.npmjs.com/package/icore)
 [![license](https://img.shields.io/npm/l/icore.svg)](LICENSE)
 
-Small dependency-free command line interface and terminal presentation mechanics for [Node.js®](https://nodejs.org) applications.
+Small dependency-free command line interface and terminal presentation
+mechanics for [Node.js®](https://nodejs.org) applications.
 
-Supports a practical GNU-style option syntax:
+`icore` owns the path from `process.argv` to terminal output:
 
-- long options: `--name value`, `--name=value`;
-- boolean flags: `--flag`, `--no-flag`;
-- short aliases: `-f`, `-n value`;
-- option terminator: `--`.
+- declarative, type-inferred option schemas;
+- multi-segment command resolution;
+- separate prepare and execute phases;
+- JSON, CSV, and text-table presentation;
+- stdout/stderr writers and reusable terminal error policy.
 
-## Installation
+Application-specific API calls, configuration, domain behavior, and help text
+remain in the consuming application.
 
-To use `icore` in your project, run:
+## Requirements And Installation
+
+`icore` requires Node.js `>=16.9.0` and includes TypeScript declarations.
 
 ```sh
 npm install icore
 ```
 
-## Table of Contents
-
-- [API Reference](#api-reference)
-  - [`createTerminalApp()`](#createterminalapp)
-  - [`createCommand()`](#createcommand)
-    - [`command.define(command)`](#commanddefinecommand)
-    - [`command.registry(commands)`](#commandregistrycommands)
-    - [`command.run(command, args, context)`](#commandruncommand-args-context)
-    - [`createCommands(commands)`](#createcommandscommands)
-    - [`commands.prepare(args, options?)`](#commandsprepareargs-options)
-    - [`commands.run(prepared, context)`](#commandsrunprepared-context)
-    - [`commands.runFromArgs(args, context, options?)`](#commandsrunfromargsargs-context-options)
-  - [`createPresentation()`](#createpresentation)
-    - [`presentation.empty()`](#presentationempty)
-    - [`presentation.text(value)`](#presentationtextvalue)
-    - [`presentation.record(value)`](#presentationrecordvalue)
-    - [`presentation.records(values)`](#presentationrecordsvalues)
-    - [`presentation.table(rows)`](#presentationtablerows)
-    - [`presentation.csv(rows)`](#presentationcsvrows)
-    - [`presentation.render(result, format?)`](#presentationrenderresult-format)
-    - [`presentation.renderers.*`](#presentationrenderers)
-  - [`createOutput()`](#createoutput)
-    - [`output.write(chunk)`](#outputwritechunk)
-    - [`output.error(chunk)`](#outputerrorchunk)
-    - [`output.stdout.write(chunk)`](#outputstdoutwritechunk)
-    - [`output.stderr.write(chunk)`](#outputstderrwritechunk)
-  - [Lower-Level Mechanics](#lower-level-mechanics)
-    - [`parseArgv(args, schema?)`](#parseargv)
-    - [`parseOptionsDetailed(schema, values)`](#parseoptionsdetailed)
-    - [`resolveCommand(registry, positionals)`](#resolvecommand)
-    - [`resolveCommandFromArgs(registry, args)`](#resolvecommandfromargs)
-    - [`prepareCommandFromArgs(registry, args, options?)`](#preparecommandfromargs)
-    - [`runPreparedCommand(prepared, context)`](#runpreparedcommand)
-    - [`runCommandFromRegistry(registry, args, context, options?)`](#runcommandfromregistry)
-    - [`runCommand(command, args, context, options?)`](#runcommand)
-- [Internal Source Layout](#internal-source-layout)
-- [How It Works](#how-it-works)
-- [Example](#example)
-- [Option Schemas](#option-schemas)
-- [Type Inference](#type-inference)
-- [Facade of arguments](#facade-of-arguments)
-- [Error Messages](#error-messages)
-- [Project Boundary](#project-boundary)
-
-## API Reference
-
-Start at `createTerminalApp()` for regular terminal applications, then move to
-the construction methods only when the app needs explicit command,
-presentation, or output wiring.
-
-### `createTerminalApp()`
-
-`createTerminalApp()` is the top-level terminal application composition point.
-It wires command execution, presentation rendering, and stdout/stderr delivery
-without taking ownership of application behavior.
-
-The checked TypeScript contract lives in [`src/terminal/app.ts`](src/terminal/app.ts).
+Import public APIs from `icore`. Deep imports into `dist` or `src` are not part
+of the package contract.
 
-The method returns an application object with `prepare(args)`,
-`writePreparedOutput(prepared, output)`, `runPrepared(prepared, context)`, and
-`run(args, context)`.
+## Contents
 
-Construction inputs:
-
-- `commands` is required; build it with [`createCommand()`](#createcommand) and
-  [`command.registry(commands)`](#commandregistrycommands), or directly with
-  [`createCommands(commands)`](#createcommandscommands);
-- `presentation` is optional; omit it to use [`createPresentation()`](#createpresentation);
-- `output` is optional; omit it to use [`createOutput()`](#createoutput);
-- `resolveFormat` is optional and customizes how a prepared command selects a
-  presentation format.
-
-Returned app methods:
+- [Quick Start](#quick-start)
+- [Choose An API Level](#choose-an-api-level)
+- [Supported Argument Syntax](#supported-argument-syntax)
+- [Error Handling](#error-handling)
+- [Guides](#guides)
+- [Project Boundary](#project-boundary)
 
-- `app.prepare(args, options?)` delegates to
-  [`commands.prepare(args, options?)`](#commandsprepareargs-options);
-- `app.writePreparedOutput(prepared, output)` renders and writes already
-  obtained terminal output without running the command;
-- `app.runPrepared(prepared, context)` runs an already prepared command, then
-  renders and writes terminal output;
-- `app.run(args, context, options?)` prepares the command, delegates command
-  execution to [`commands.run(prepared, context)`](#commandsrunprepared-context),
-  then renders and writes terminal output.
+## Quick Start
 
-Minimal shape:
+Create commands, put them in a registry, and give that registry to
+`createTerminalApp()`:
 
 ```ts
-const app = createTerminalApp({
-  commands,
-  presentation,
-  output
-});
+import {
+  createCommand,
+  createTerminalApp
+} from 'icore';
 
-const args = process.argv.slice(2);
-const exitCode = await app.run(args, context, {
-  strict: true
-});
-```
-
-Command handlers keep ownership of application work. Terminal output methods
-accept terminal-ready results: text, streaming text, presentation results, or no
-output. If a command returns an application lifecycle object, run it with
-`app.commands.run(...)`, inspect or map the result, then pass only terminal
-output to `app.writePreparedOutput(...)`. Application DTO mapping, config
-loading, network clients, and domain behavior stay in the consuming application.
-
-String output is written exactly as provided. Return `\n` from the command when
-line output is desired.
-
-### `createCommand()`
-
-`createCommand()` returns the command mechanics entrypoint used to build
-`commands` for `createTerminalApp()`.
-
-Simplified shape:
-
-```ts
 const command = createCommand();
 
-command.define(commandDefinition);
-command.registry(commandDefinitions);
-command.run(commandDefinition, args, context);
-```
-
-Build `commands` before creating the terminal app:
-
-- `createCommand()` creates the command mechanics entrypoint;
-- `command.define(...)` declares one command;
-- `command.registry([...])` returns the `commands` object required by
-  `createTerminalApp()`.
-
-The command definition keeps the application-specific parts: command path,
-option schema, and handler behavior. The terminal app only needs the resulting
-registry object.
-
-#### `command.define(command)`
-
-Declares one command and preserves its literal path and option schema types.
-Use it when defining commands inline before adding them to a registry.
-
-#### `command.registry(commands)`
-
-Builds the `commands` object required by `createTerminalApp()`. It keeps the
-registered command definitions, derived command names, and command flow methods
-together.
-
-#### `command.run(command, args, context)`
-
-Runs a single command without a registry. Use it for focused command execution,
-small tests, or custom flows where command path resolution is not needed.
-
-#### `createCommands(commands)`
-
-Use `createCommands(commands)` when the application already has command
-definitions and does not need the `createCommand()` object form.
-
-The returned object is the same `commands` contract consumed by
-`createTerminalApp()`: it can resolve, prepare, and run registered commands.
-
-Simplified shape:
-
-```ts
-const commands = createCommands(commandDefinitions);
-
-commands.names;
-commands.registry;
-commands.resolve(positionals);
-commands.resolveFromArgs(args);
-commands.prepare(args, options);
-commands.run(prepared, context);
-commands.runFromArgs(args, context, options);
-```
-
-Use `commands.prepare(...)` and `commands.run(...)` when application code needs
-the same two-phase flow used by `createTerminalApp()`. Use
-`commands.runFromArgs(...)` for custom terminal boundaries that still want the
-registry-level command flow.
-
-#### `commands.prepare(args, options?)`
-
-Resolves and validates a registered command without runtime context. This is
-the same preparation step used by `app.prepare(...)` and `app.run(...)`.
-
-#### `commands.run(prepared, context)`
-
-Runs an already prepared command with application context. Use it after
-`commands.prepare(...)` when the application needs a custom two-phase flow.
-
-#### `commands.runFromArgs(args, context, options?)`
-
-Resolves, prepares, and runs a registered command without the terminal
-application boundary. Use it when a custom boundary owns presentation or output.
-
-### `createPresentation()`
-
-`createPresentation()` creates the presentation object used by
-`createTerminalApp()` to render terminal-ready results.
-
-Source of truth:
-
-- [src/presentation/facade.ts](src/presentation/facade.ts) for the presentation
-  object;
-- [src/presentation/view.ts](src/presentation/view.ts) for presentation view
-  factories;
-- [src/presentation/result-renderer.ts](src/presentation/result-renderer.ts)
-  for rendering presentation results.
-
-Simplified shape:
-
-```ts
-const presentation = createPresentation();
-
-presentation.empty();
-presentation.text(value);
-presentation.record(value);
-presentation.records(values);
-presentation.table(rows);
-presentation.csv(rows);
-presentation.render(result, format);
-```
-
-The presentation object owns generic JSON, CSV, and table rendering mechanics.
-Application code still maps domain objects to presentation-ready values.
-
-#### `presentation.empty()`
-
-Creates a presentation result with no terminal output.
-
-#### `presentation.text(value)`
-
-Wraps ready terminal text. Use it when the command already owns the final text.
-
-#### `presentation.record(value)`
-
-Creates a generic one-record view. The renderer can print it as table, CSV, or
-JSON depending on the selected format.
-
-#### `presentation.records(values)`
-
-Creates a generic multi-record view. Application code still decides which fields
-belong in each record.
-
-#### `presentation.table(rows)`
-
-Creates an explicit table view from prepared text rows.
-
-#### `presentation.csv(rows)`
-
-Creates an explicit CSV view from scalar rows.
-
-#### `presentation.render(result, format?)`
-
-Renders a presentation result to terminal text. The default format is `table`.
-
-#### `presentation.renderers.*`
-
-Exposes lower-level JSON, table, and CSV renderers for custom presentation
-composition. Prefer `presentation.render(...)` for regular terminal commands.
-
-### `createOutput()`
-
-`createOutput()` creates the output object used by `createTerminalApp()` to
-write rendered text.
-
-Source of truth:
-
-- [src/output/facade.ts](src/output/facade.ts) for the output object;
-- [src/output/node-writer.ts](src/output/node-writer.ts) for stdout and stderr
-  writers;
-- [src/output/text-writer.ts](src/output/text-writer.ts) for backpressure-aware
-  text writing.
-
-Simplified shape:
-
-```ts
-const output = createOutput();
-
-await output.write(chunk);
-await output.error(chunk);
-await output.stdout.write(chunk);
-await output.stderr.write(chunk);
-```
-
-The default output writes regular text to `stdout` and diagnostic text to
-`stderr`. Pass custom sinks when tests or applications need controlled output.
-
-#### `output.write(chunk)`
-
-Writes regular terminal output through stdout.
-
-#### `output.error(chunk)`
-
-Writes diagnostic terminal output through stderr.
-
-#### `output.stdout.write(chunk)`
-
-Writes directly to the stdout channel. Use this only when a specific channel
-must be passed around.
-
-#### `output.stderr.write(chunk)`
-
-Writes directly to the stderr channel. Use this only when a specific channel
-must be passed around.
-
-### Lower-Level Mechanics
-
-Lower-level mechanics sit behind the construction methods. They are kept out of
-the main navigation to keep the application construction path first.
-
-For application code, prefer `createTerminalApp()` first. Drop to this level
-only when the terminal application flow or construction methods are too coarse.
-
-Source of truth:
-
-- [src/argv/parser.ts](src/argv/parser.ts) for raw argv parsing;
-- [src/options/parser.ts](src/options/parser.ts) for option value validation;
-- [src/command/mechanics.ts](src/command/mechanics.ts) for command resolution,
-  preparation, and execution.
-
-#### Parsing And Resolution
-
-This group turns raw arguments or positionals into a selected command.
-
-- <a id="parseargv"></a>`parseArgv(args, schema?)` parses raw CLI arguments into positionals and raw
-  option values;
-- <a id="resolvecommand"></a>`resolveCommand(registry, positionals)` resolves from already parsed
-  positionals;
-- <a id="resolvecommandfromargs"></a>`resolveCommandFromArgs(registry, args)` resolves from raw CLI arguments by
-  using registered command schemas.
-
-Use these methods when command selection is needed without validation or
-execution.
-
-#### Preparation
-
-This group validates command input without runtime context.
-
-- <a id="parseoptionsdetailed"></a>`parseOptionsDetailed(schema, values)` validates raw option values and keeps
-  option presence metadata;
-- <a id="preparecommandfromargs"></a>`prepareCommandFromArgs(registry, args, options?)` resolves and validates a
-  registered command;
-- `options.strict` rejects options before the command path.
-
-Use preparation when runtime resources should be created only after command
-selection is known.
-
-#### Execution
-
-This group calls command handlers after command input is prepared.
-
-- <a id="runpreparedcommand"></a>`runPreparedCommand(prepared, context)` runs an already prepared command;
-- <a id="runcommandfromregistry"></a>`runCommandFromRegistry(registry, args, context, options?)` prepares and runs
-  a command from a registry;
-- <a id="runcommand"></a>`runCommand(command, args, context, options?)` runs a single command without a
-  registry.
-
-Use execution primitives for custom boundaries. Regular CLI applications should
-usually stay at `createTerminalApp()`.
-
-## Internal Source Layout
-
-The public package entrypoint remains `src/index.ts`. Internal source files are
-grouped by CLI framework responsibility:
-
-```text
-src/
-  argv/          raw argv token parsing
-  options/       option schemas and option value validation
-  command/       command resolution and execution mechanics
-  presentation/  terminal view models and JSON/CSV/table renderers
-  output/        stdout/stderr writer boundaries
-  terminal/      command + presentation + output app composition
-  errors/        machine-readable icore errors
-```
-
-Consumers should continue importing from `icore`; deep imports are an internal
-source layout detail.
-
-## How It Works
-
-![yuml diagram](http://yuml.me/diagram/scruffy;dir:LR;/class/[*argv*%20{bg:gray}|External;hello%20--name%20Alice%20--uppercase]->[*matches*%20{bg:lavender}|System;parse,%20resolve,%20validate,%20infer]->[*typed%20result*%20{bg:honeydew}|Container;command=hello;%20name=Alice;%20uppercase=true]->[*your%20app*%20{bg:cornsilk}|System;business%20logic%20and%20output])
-
-## Example
-
-More examples live in [examples/readme.md](examples/readme.md), including
-option schemas, CLI argument syntax, practical CLI patterns, and lower-level
-primitives.
-
-```ts
-import { defineCommand, runCommand } from 'icore';
-
-const exampleCommand = defineCommand({
-  path: ['hello'],
-  options: {
-    name: {
-      type: 'string',
-      default: 'world'
+const commands = command.registry([
+  command.define({
+    path: ['hello'],
+    options: {
+      name: {
+        type: 'string',
+        default: 'world'
+      },
+      uppercase: {
+        type: 'boolean'
+      }
     },
-    uppercase: {
-      type: 'boolean'
+    handle({ options }) {
+      const greeting = `Hello, ${options.name}!`;
+
+      return `${options.uppercase ? greeting.toUpperCase() : greeting}\n`;
     }
-  },
-  async handle({ options }) {
-    const message = `Hello, ${options.name}!`;
+  })
+] as const);
 
-    return options.uppercase ? message.toUpperCase() : message;
-  }
-});
+const app = createTerminalApp({ commands });
 
-const output = await runCommand(
-  exampleCommand,
-  ['hello', '--name', 'Alice', '--uppercase'],
-  {}
-);
+async function main(args: readonly string[]): Promise<void> {
+  process.exitCode = await app.run(args, undefined, {
+    strict: true
+  });
+}
 
-console.log(output);
+void main(process.argv.slice(2));
 ```
 
-Terminal output:
+After compiling the application:
 
 ```console
-$ node cli.js hello --name Alice --uppercase
-HELLO, ALICE!
-```
-
-The command handler receives parsed options, user-provided option metadata,
-remaining positionals, prepared payload, and caller provided context.
-
-## Option Schemas
-
-Options are described as plain objects.
-
-**Option names are exact.** `icore` does not normalize `camelCase` to
-`kebab-case`. Use quoted object keys when your public CLI option contains
-`-`.
-
-Each option can define an optional short alias:
-
-```ts
-const schema = {
-  name: {
-    type: 'string',
-    alias: 'n'
-  },
-  uppercase: {
-    type: 'boolean',
-    alias: 'u'
-  }
-} as const;
-```
-
-Aliases must be a single ASCII letter and unique within the schema. Parsed
-values are always returned by long option name.
-
-### `type: 'string'`
-
-```ts
-const schema = {
-  name: {
-    type: 'string',
-    required: true
-  },
-  style: {
-    type: 'string',
-    choices: ['short', 'long'],
-    default: 'short'
-  }
-} as const;
-```
-
-String options reject missing required values, blank strings such as `--name=`,
-boolean flag form, and values outside `choices`.
-
-### `type: 'boolean'`
-
-```ts
-const schema = {
-  uppercase: {
-    type: 'boolean'
-  }
-} as const;
-```
+$ node dist/cli.js hello
+Hello, world!
 
-Boolean options accept **flag form** and schema-known negation:
-
-```sh
---uppercase
---no-uppercase
-```
-
-Explicit values are rejected:
-
-```sh
---uppercase=true
---uppercase=false
---uppercase=yes
---uppercase=
-```
-
-`--uppercase false` keeps `--uppercase` as `true` and leaves `false` as a positional
-argument.
-
-Use `syntax: 'flag'` when a boolean option should accept only flag form:
-
-```ts
-const schema = {
-  uppercase: {
-    type: 'boolean',
-    default: false,
-    syntax: 'flag'
-  }
-} as const;
+$ node dist/cli.js hello --name Alice --uppercase
+HELLO, ALICE!
 ```
 
-With `syntax: 'flag'`, `--uppercase` is accepted, while `--uppercase=value`
-and `--no-uppercase` are rejected.
+The schema determines the handler's option types. Required options and options
+with defaults are always present; optional options are returned as
+`T | undefined`.
 
-### `type: 'number'`
+`strict: true` rejects options placed before the command path. String output is
+written exactly as returned, so include `\n` when line output is intended.
 
-```ts
-const schema = {
-  limit: {
-    type: 'number',
-    integer: true,
-    min: 1,
-    max: 1000,
-    default: 100
-  }
-} as const;
-```
+## Choose An API Level
 
-Number options parse decimal numeric values and can validate integer and range
-constraints. Defaults are validated with the same rules as user-provided values.
+Start with the highest-level API that fits the application:
 
-## Type Inference
+| Need | Start with | Detailed guide |
+| --- | --- | --- |
+| Complete terminal application | `createTerminalApp()` | [Terminal App](examples/terminal-app.md) |
+| Commands and option schemas | `createCommand()` or `createCommands()` | [Option Schemas](examples/option-schemas.md) |
+| Custom prepare/execute lifecycle | `commands.prepare()` and `commands.run()` | [Custom Command Flow](examples/custom-command-flow.md) |
+| Presentation without command execution | `createPresentation()` | [Presentation And Output](examples/presentation-output.md) |
+| Explicit stdout/stderr writing | `createOutput()` | [Output Writers](examples/output-writers.md) |
+| Parser and resolver primitives | `parseArgv()`, `resolveCommand()`, and related exports | [Primitive Mechanics](examples/readme.md#primitive-mechanics) |
 
-Use `InferOptions` when you need the parsed option type explicitly.
+The terminal application exposes the main lifecycle operations:
 
-```ts
-import type { InferOptions } from 'icore';
-
-const schema = {
-  name: {
-    type: 'string',
-    default: 'world'
-  },
-  uppercase: {
-    type: 'boolean'
-  }
-} as const;
+- `app.run(args, context, options?)` prepares, executes, renders, and writes;
+- `app.prepare(args, options?)` validates input without runtime context;
+- `app.runPrepared(prepared, context)` executes an already prepared command;
+- `app.writePreparedOutput(prepared, output)` writes caller-obtained output;
+- `app.reportError(error, context?)` applies the configured terminal error
+  policy and returns a process-style exit code.
 
-type Options = InferOptions<typeof schema>;
-```
+Command handlers receive parsed `options`, option-presence metadata in
+`provided`, remaining `positionals`, caller-owned `context`, and an optional
+prepared `payload`.
 
-`Options` is equivalent to:
+Supported terminal results are strings, async string streams, presentation
+results, and `undefined`:
 
-```ts
-type Options = {
-  name: string;
-  uppercase: boolean | undefined;
-};
+```text
+argv → resolve → validate/prepare → execute → render → stdout/stderr
 ```
 
-**Required options and options with defaults are always present.** Optional
-options without defaults are returned as `T | undefined`.
-
-Use `InferProvidedOptions` when you need the option presence type explicitly.
-
-```ts
-import type { InferProvidedOptions } from 'icore';
+Exact public exports live in [`src/index.ts`](src/index.ts) and in the bundled
+TypeScript declarations.
 
-type Provided = InferProvidedOptions<typeof schema>;
-```
+## Supported Argument Syntax
 
-`Provided` maps every schema option to `boolean`. `true` means the user
-specified that option explicitly; defaults keep the flag `false`.
+The supported syntax is intentionally small and predictable:
 
-Use `MergeOptionsSchemas` when you need the merged schema type explicitly.
+| Form | Example | Notes |
+| --- | --- | --- |
+| Long option with separate value | `--name Alice` | String and number options consume a value |
+| Long option with attached value | `--name=Alice` | Equivalent to the separate form |
+| Boolean flag | `--verbose` | Produces `true` |
+| Boolean negation | `--no-cache` | Supported for known boolean options unless `syntax: 'flag'` |
+| Short alias | `-v`, `-n Alice` | Must be declared in the schema as one ASCII letter |
+| Option terminator | `--` | Every following token becomes positional |
 
-```ts
-import type { MergeOptionsSchemas } from 'icore';
+Option names are exact: `icore` does not convert `camelCase` to `kebab-case`.
+Explicit boolean values such as `--verbose=true`, attached short values such as
+`-nAlice`, and grouped aliases such as `-abc` are not supported.
 
-type Schema = MergeOptionsSchemas<[typeof nameOptions, typeof greetingOptions]>;
-```
+See [CLI Argument Syntax](examples/cli-argument-syntax.md) for parsing examples,
+edge cases, duplicate handling, and the option terminator contract.
 
-Use `CommandName` when you need the inferred command name type explicitly.
+## Error Handling
 
-```ts
-import type { CommandName } from 'icore';
+CLI parsing, validation, resolution, and definition failures are reported as
+`IcoreError`. It extends `Error` with:
 
-type Name = CommandName<typeof helloFormalCommand>;
-```
+- a stable machine-readable `code`;
+- a `usage` or `definition` category;
+- required, code-specific `details`.
 
-`Name` is equivalent to:
+Use `isIcoreError(...)` instead of manually casting details. Passing a code to
+the guard narrows the corresponding details shape:
 
 ```ts
-type Name = 'hello formal';
-```
+import {
+  createTerminalApp,
+  isIcoreError
+} from 'icore';
 
-## Facade of arguments
+const help = 'Usage: cli hello [--name value]';
 
-Use `--` to stop option parsing. The terminator itself is not included in
-positionals; every following token is treated as positional, even when it starts
-with `-`.
-
-Short syntax is supported only for aliases declared in the option schema.
-Boolean aliases use flag form, such as `-f`; string and number aliases use a
-separate value, such as `-n value`.
-
-Attached short values such as `-nvalue` and grouped short booleans such as
-`-abc` are not supported yet. Unknown short tokens remain positional for
-compatibility.
-
-Negated syntax such as `--no-cache` is interpreted as `cache: false` when
-`cache` is a known boolean option without `syntax: 'flag'`. Unknown negated
-options, negation for string or number options, and negation for flag-only
-boolean options are rejected.
-
-## Error Messages
-
-`icore` throws `IcoreError` objects for CLI parsing, option validation, and
-command resolution failures. `IcoreError` extends the regular `Error` class and
-adds a stable machine-readable `code` plus structured `details`.
+const app = createTerminalApp({
+  commands,
+  errorPolicy: {
+    renderError(error) {
+      const message = error instanceof Error
+        ? error.message
+        : String(error);
 
-Applications should treat `error.message` as **display text**. Use `error.code`
-for machine-readable handling:
+      if (isIcoreError(error, 'UNKNOWN_COMMAND')) {
+        return `${message}\n\n${help}\n`;
+      }
 
-```ts
-import { IcoreError } from 'icore';
-
-try {
-  await main(args);
-} catch (error) {
-  if (error instanceof IcoreError && error.code === 'UNKNOWN_COMMAND') {
-    printHelp();
-    process.exitCode = 2;
-    return;
+      return `${message}\n`;
+    },
+    resolveExitCode(error) {
+      return isIcoreError(error) && error.category === 'usage'
+        ? 2
+        : 1;
+    }
   }
-
-  throw error;
-}
+});
 ```
 
-Supported error codes:
+Inside the `UNKNOWN_COMMAND` branch, fields such as `error.details.command` and
+`error.details.positionals` are strongly typed. `IcoreErrorDetailsMap` is the
+public source of truth for every code. Direct `new IcoreError(...)` calls must
+provide a third argument matching the selected code.
 
-```ts
-type IcoreErrorCode =
-  | 'UNKNOWN_COMMAND'
-  | 'UNEXPECTED_ARGUMENT'
-  | 'DUPLICATE_ARGUMENT'
-  | 'EXPECTED_REQUIRED_ARGUMENT'
-  | 'INVALID_OPTION_TYPE'
-  | 'INVALID_OPTION_CHOICE'
-  | 'UNEXPECTED_POSITIONAL'
-  | 'INVALID_OPTION_ALIAS'
-  | 'DUPLICATE_ALIAS'
-  | 'INVALID_OPTION_DEFAULT'
-  | 'DUPLICATE_COMMAND';
-```
+Without a custom policy, terminal apps write `Error.message + "\n"` (or
+`String(error) + "\n"` for other thrown values) and return exit code `1`.
+Application-specific help remains application policy.
 
-Applications can catch these errors and decide how to print them. For example,
-after printing `error.message`, terminal output can look like this:
+Custom lifecycles can call `app.reportError(...)` to reuse the same rendering
+and exit-code policy. The complete prepare, execute, write, and external phase
+flow is shown in the
+[Terminal App guide](examples/terminal-app.md#reuse-error-reporting-in-a-custom-lifecycle).
 
-```console
-$ node cli.js hello --unknown
-Unexpected argument '--unknown'
+## Guides
 
-$ node cli.js hello --uppercase=yes
-Expected '--uppercase' as boolean flag
+The [examples index](examples/readme.md) routes from regular terminal apps to
+lower-level mechanics. Useful starting points:
 
-$ node cli.js hello --name=
-Expected '--name' as string
-```
+- [Terminal App](examples/terminal-app.md) — regular application composition,
+  format resolution, and reusable error reporting;
+- [Option Schemas](examples/option-schemas.md) — strings, booleans, numbers,
+  choices, defaults, aliases, and inferred types;
+- [Practical CLI Patterns](examples/practical-cli-patterns.md) — help/version
+  shortcuts, shared options, and compatibility aliases;
+- [Command Resolution](examples/command-resolution.md) — registries, explicit
+  resolution, and command-name guards;
+- [Two-Phase Primitives](examples/two-phase-primitives.md) — preparation,
+  payloads, execution, and provided-option metadata;
+- [Presentation Primitives](examples/presentation-primitives.md) — text, record,
+  table, CSV, JSON, and direct renderers.
 
-Errors thrown by command `prepare` and `handle` functions are application errors
-and pass through unchanged.
+Release history is recorded in [CHANGELOG.md](CHANGELOG.md). Directional
+decisions live in [docs/roadmap.md](docs/roadmap.md).
 
 ## Project Boundary
 
-`icore` is intended to be a **small CLI mechanics module**. It should **not**
-grow into a domain-specific framework for a particular SDK or API.
-
-Good responsibilities for `icore`:
+`icore` is a small terminal mechanics module. It owns generic behavior such as
+option validation, command resolution, typed handler input, presentation
+rendering, error contracts, and stdout/stderr delivery.
 
-- option schema evaluation;
-- command path checking;
-- common argument errors;
-- typed command handler input;
-- generic JSON, CSV, and text table rendering;
-- stdout/stderr text writer mechanics.
+It does not own application DTO mapping, API calls, configuration loading,
+domain-specific validation, lifecycle resources, or help content. Keep those
+responsibilities in the consuming application.
 
-Application-specific report mapping, scalar formatting, API calls, config
-loading, and domain behavior stay outside `icore`.
+Licensed under the [MIT License](LICENSE).