Epic P7. Locks in the epic's gains with ESLint gates so the adoption work does not regress. - ccs/no-new-throw-error (error): custom flat-config rule that flags NEW throw new Error(...) outside a generated baseline allowlist. Forces the typed-error taxonomy (src/errors/error-types.ts). Existing 338 sites are grandfathered in eslint-rules/throw-error-baseline.json; only NEW violations error. Rule normalizes the filename to repo-root-relative to match baseline keys regardless of how ESLint reports paths. - max-lines (warn, 400, skipBlankLines/skipComments): warns on files over 400 LOC (goal of P5/P6 god-file splits). Currently 51 warnings on the not-yet-split god-files (P6 territory). - scripts/generate-throw-error-baseline.js: emits the allowlist from raw source (superset of real throws; never undercounts). Run after intentionally grandfathering a site, or quarterly to prune. - tests/unit/eslint-rules/no-throw-new-error.test.ts: rule logic (flags off-allowlist, passes on-allowlist/typed/rethrow, line-sensitivity). - docs/code-standards.md: Lint Enforcement Gates section. - docs/logging-contract.md: error.code -> ExitCode table (from P4). validate + validate:ci-parity green.
21 KiB
CCS Code Standards
Last Updated: 2026-04-07
Code standards, modularization patterns, and conventions for the CCS codebase.
Core Principles
YAGNI (You Aren't Gonna Need It)
- No features "just in case"
- Only implement what is currently needed
- Delete unused code rather than commenting it out
KISS (Keep It Simple, Stupid)
- Prefer simple solutions over clever ones
- Reduce complexity at every opportunity
- Use established patterns over custom implementations
DRY (Don't Repeat Yourself)
- One source of truth for configuration
- Extract common logic into shared utilities
- Use barrel exports to centralize imports
File Organization
Directory Structure Rules
- Domain-based organization: Group files by business domain, not by file type
- Barrel exports required: Every directory must have an
index.tsaggregating exports - Flat within depth: Keep nesting to 3 levels maximum
- Co-location: Keep related files together (component + hooks + utils)
File Naming Conventions
| Convention | Example | When to Use |
|---|---|---|
| kebab-case | cliproxy-executor.ts |
All TypeScript/TSX files |
| kebab-case | profile-detector.ts |
Multi-word file names |
| *-adapter.ts | claude-adapter.ts, droid-adapter.ts |
TargetAdapter implementations |
| *-detector.ts | droid-detector.ts |
Binary detection logic |
| *-manager.ts | droid-config-manager.ts |
Config/state management |
| PascalCase | BinaryManager |
Class exports only |
| camelCase | detectProfile |
Function exports |
File names should be descriptive: LLMs should understand the file's purpose from its name alone without reading content.
Correct Examples
src/cliproxy/binary-manager.ts # Binary management logic
src/commands/doctor-command.ts # Doctor CLI command handler
ui/src/components/cliproxy/provider-editor/index.tsx
Incorrect Examples
src/utils/helper.ts # Too vague
src/cliproxy/manager.ts # Which manager?
ui/src/components/Editor.tsx # Not kebab-case
File Size Limit: 200 Lines
Target: All code files should be under 200 lines.
Exceptions (with justification):
- Data files (model-pricing.ts, model-catalog.ts)
- Entry points with routing logic (ccs.ts)
- Complex transformation logic that cannot be meaningfully split
Why 200 Lines?
- Context efficiency: LLMs process smaller files faster
- Single responsibility: Forces focused, testable modules
- Navigation: Easier to scan and understand
- Maintainability: Reduces merge conflicts
When Files Exceed 200 Lines
If a file grows beyond 200 lines:
-
Identify extraction candidates:
- Helper functions that could be utilities
- Constants and type definitions
- Subcomponents within React components
- Related logic that forms a cohesive unit
-
Create subdirectory structure:
# Before provider-editor.tsx (921 lines) # After provider-editor/ ├── index.tsx # Main component (200 lines) ├── model-mapping-form.tsx ├── endpoint-config.tsx ├── auth-section.tsx ├── hooks.ts ├── types.ts └── utils.ts -
Preserve public API: Main export remains the same through barrel export
Barrel Export Pattern
What is a Barrel Export?
An index.ts file that aggregates and re-exports module contents:
// src/cliproxy/index.ts
// Types (with explicit type keyword)
export type { PlatformInfo, BinaryInfo } from './types';
// Functions
export { detectPlatform } from './platform-detector';
export { BinaryManager } from './binary-manager';
// From subdirectories
export * from './auth';
export * from './services';
Rules for Barrel Exports
- Every domain directory must have
index.ts - Export types with
export typefor tree-shaking - Re-export subdirectories for deep access
- Keep barrel exports flat - no logic, only exports
Import Patterns
// CORRECT: Import from domain barrel
import { execClaudeWithCLIProxy, CLIProxyProvider } from '../cliproxy';
import { Config, Settings } from '../types';
// INCORRECT: Import from specific file (bypasses barrel)
import { execClaudeWithCLIProxy } from '../cliproxy/cliproxy-executor';
Exception: Deep Imports
Allowed when:
- Importing private utilities not exposed in barrel
- Circular dependency avoidance
- Performance-critical tree-shaking
Target Adapter Pattern
The target adapter pattern enables pluggable support for multiple CLI implementations (Claude Code, Factory Droid, Codex CLI, etc.) while preserving a unified profile system.
Pattern Overview
Each CLI target implements a TargetAdapter interface:
interface TargetAdapter {
readonly type: TargetType; // 'claude' | 'droid' | 'codex'
readonly displayName: string; // Human-readable name
detectBinary(): TargetBinaryInfo | null; // Find CLI on system
prepareCredentials(creds: TargetCredentials): Promise<void>; // Deliver credentials
buildArgs(profile: string, userArgs: string[]): string[]; // Build CLI args
buildEnv(creds: TargetCredentials, type: string): Env; // Build env vars
exec(args: string[], env: Env): void; // Spawn CLI process
supportsProfileType(type: string): boolean; // Validate profile
}
Key Differences Per Target
| Aspect | Claude | Droid | Codex |
|---|---|---|---|
| Credential delivery | Environment variables | Config file (~/.factory/settings.json) | Transient -c overrides + CCS_CODEX_API_KEY |
| Spawn args | claude <args> |
droid -m custom:ccs-<profile> <args> |
codex <args> or codex -c ... <args> |
| Config write | None (uses env) | upsertCcsModel() writes to settings |
None at runtime; dashboard edits user-owned ~/.codex/config.toml only |
| Binary detection | detectClaudeCli() |
detectDroidCli() with version check |
detectCodexCli() plus --config capability probe |
Target Resolution Priority
Resolves which adapter to use via resolveTargetType():
1. --target <name> flag (highest priority)
↓
2. explicit runtime entrypoint (`CCS_INTERNAL_ENTRY_TARGET`):
- ccs-droid / ccsd → droid
- ccs-codex / ccsx → codex
- ccsxp → codex (native cliproxy shortcut)
↓
3. argv[0] detection (runtime alias pattern / custom alias map):
- ccs-droid → droid
- ccsd → droid
- ccs-codex → codex
- ccsx → codex
- ccs → default
↓
4. Profile config: profileConfig.target field
↓
5. Fallback: 'claude' (lowest priority)
Registration Pattern
At startup, adapters self-register into the runtime registry:
// In ccs.ts or initialization
registerTarget(new ClaudeAdapter());
registerTarget(new DroidAdapter());
registerTarget(new CodexAdapter());
// Later, when executing
const targetType = resolveTargetType(args, profileConfig);
const adapter = getTarget(targetType);
await adapter.prepareCredentials(credentials);
const spawnArgs = adapter.buildArgs(profile, userArgs);
adapter.exec(spawnArgs, adapter.buildEnv(credentials, profileType));
Adding a New Target
To add support for a new CLI (e.g., newcli):
- Create
src/targets/newcli-adapter.tsimplementingTargetAdapter - Implement each required method (detection, credential delivery, spawning)
- Create
src/targets/newcli-detector.tsfor binary detection logic - Export from
src/targets/index.ts - Register in
ccs.ts:registerTarget(new NewCliAdapter()) - Update
TargetTypeunion to include'newcli'
Monster File Splitting Methodology
When splitting large files (500+ lines), follow this process:
Step 1: Analyze Structure
Identify logical boundaries:
- Render sections in React components
- Handler groups in route files
- Related utility functions
- Constants and types
Step 2: Extract Types First
// types.ts
export interface ProviderEditorProps {
providerId: string;
onSave: (config: ProviderConfig) => void;
}
export interface ModelMappingValues {
model: string;
endpoint: string;
}
Step 3: Extract Utilities
// utils.ts
export function validateEndpoint(url: string): boolean { ... }
export function formatModelName(name: string): string { ... }
Step 4: Extract Hooks
// hooks.ts
export function useProviderConfig(providerId: string) { ... }
export function useModelValidation() { ... }
Step 5: Extract Subcomponents
// model-mapping-form.tsx
export function ModelMappingForm({ values, onChange }: Props) { ... }
Step 6: Compose in Index
// index.tsx
import { ModelMappingForm } from './model-mapping-form';
import { useProviderConfig } from './hooks';
import type { ProviderEditorProps } from './types';
export function ProviderEditor({ providerId, onSave }: ProviderEditorProps) {
const config = useProviderConfig(providerId);
return (
<div>
<ModelMappingForm values={config.mapping} onChange={...} />
</div>
);
}
// Re-export types for consumers
export type { ProviderEditorProps, ModelMappingValues } from './types';
TypeScript Standards
Strict Mode Required
All projects use TypeScript strict mode:
{
"compilerOptions": {
"strict": true,
"noUnusedLocals": true,
"noUnusedParameters": true,
"noImplicitReturns": true,
"noFallthroughCasesInSwitch": true
}
}
Type Annotations
// CORRECT: Explicit return types for public functions
export function detectProfile(args: string[]): DetectedProfile { ... }
// CORRECT: Inferred types for internal functions
const formatName = (name: string) => name.trim().toLowerCase();
// INCORRECT: any type
function processData(data: any) { ... } // Use unknown or proper type
Type Exports
// CORRECT: Use type keyword for type-only exports
export type { Config, Settings } from './config';
// CORRECT: Group type exports in barrel
export type {
PlatformInfo,
BinaryInfo,
DownloadProgress,
} from './types';
ESLint Rules (Enforced)
| Rule | Level | Notes |
|---|---|---|
@typescript-eslint/no-unused-vars |
error | Ignore _ prefix |
@typescript-eslint/no-explicit-any |
error | Use proper types |
@typescript-eslint/no-non-null-assertion |
error | No ! assertions |
prefer-const |
error | Immutable by default |
no-var |
error | Use const/let |
eqeqeq |
error | Strict equality |
react-hooks/* |
recommended | (UI only) |
Terminal Output Standards
CCS Logging Standards
- Use the shared logger from
src/services/logging/for CCS-owned runtime diagnostics, request tracing, and structured events. - Keep
utils/uiand deliberateconsole.log/console.erroroutput for user-facing CLI UX only. - Redact secrets before persistence; never write raw tokens, cookies, API keys, or password hashes into CCS-owned logs.
- Persist CCS-owned logs only under
getCcsDir()/logs; do not invent per-feature log roots. - When adding dashboard polling or diagnostics routes, prevent them from recursively logging the log viewer itself.
ASCII Only
// CORRECT
console.log('[OK] Operation successful');
console.log('[!] Warning message');
console.log('[X] Error occurred');
console.log('[i] Information');
// INCORRECT - NO EMOJIS
console.log('Operation successful'); // NO
console.log('Warning message'); // NO
Color Handling
import { colors } from '../utils/ui';
// Colors are TTY-aware and respect NO_COLOR
console.log(colors.green('[OK]') + ' Operation successful');
Box Borders
Use ASCII box drawing for error displays:
+=====================================+
| [X] ERROR: Configuration failed |
| |
| Details: Unable to parse config |
+=====================================+
Cross-Platform Adapter Spawning
When implementing target adapters, handle platform differences for binary spawning:
// Window shell detection (.cmd, .bat, .ps1 require shell)
const needsShell = isWindows && /\.(cmd|bat|ps1)$/i.test(binaryPath);
if (needsShell) {
// Escape arguments and use shell: true
const cmdString = [binaryPath, ...args].map(escapeShellArg).join(' ');
spawn(cmdString, { shell: true, stdio: 'inherit' });
} else {
// Direct spawn (Unix-like, unshelled Windows executables)
spawn(binaryPath, args, { stdio: 'inherit' });
}
This pattern is used in both ClaudeAdapter and DroidAdapter to ensure cross-platform consistency.
For all Claude child-process launches (delegation, adapters, proxies, helper spawners), sanitize env before spawn:
const cleanEnv = stripClaudeCodeEnv(mergedEnv); // case-insensitive remove of CLAUDECODE
spawn(binaryPath, args, { env: cleanEnv, stdio: 'inherit' });
This prevents Claude Code nested-session guard failures when CCS runs inside parent Claude sessions.
React Component Standards (UI)
Component Structure
// component-name.tsx
// 1. Imports (grouped: react, external, internal, relative)
import { useState } from 'react';
import { Button } from '@/components/ui/button';
import { useProfiles } from '@/hooks';
import { formatName } from './utils';
import type { ComponentProps } from './types';
// 2. Types (if not in separate file)
interface Props {
id: string;
onSave: () => void;
}
// 3. Component
export function ComponentName({ id, onSave }: Props) {
// Hooks first
const profiles = useProfiles();
const [state, setState] = useState(null);
// Handlers
const handleClick = () => { ... };
// Render
return ( ... );
}
Naming Conventions
| Item | Convention | Example |
|---|---|---|
| Component files | kebab-case.tsx | provider-editor.tsx |
| Component exports | PascalCase | ProviderEditor |
| Hook files | use-*.ts | use-profiles.ts |
| Hook exports | useCamelCase | useProfiles |
| Utility files | kebab-case.ts | path-utils.ts |
| Utility exports | camelCase | formatPath |
Input State Persistence Patterns
When building forms and editors that allow users to make changes, follow these patterns to prevent data loss.
Pattern 1: Key-Based Remounting
Use when: Component has complex local state that should reset on prop changes.
// Parent component
<ProfileEditor
key={profileId} // Forces remount when profile changes
profileId={profileId}
onSave={handleSave}
/>
Why: Without key, React reuses the component instance. Local useState values persist even when props change, causing stale data bugs.
Pattern 2: Unsaved Changes Confirmation
Use when: User might navigate away while editing.
// Parent tracks dirty state
const [editorHasChanges, setEditorHasChanges] = useState(false);
const [pendingSwitch, setPendingSwitch] = useState<string | null>(null);
// Child notifies parent of dirty state
useEffect(() => {
onHasChangesUpdate?.(computedHasChanges);
}, [computedHasChanges, onHasChangesUpdate]);
// Intercept navigation
const handleSelect = (id: string) => {
if (editorHasChanges && currentId !== id) {
setPendingSwitch(id); // Show confirmation dialog
} else {
setCurrentId(id);
}
};
Flow:
- Child computes
hasChangesfrom local state vs saved data - Child notifies parent via callback
- Parent intercepts navigation when dirty
- Show confirmation dialog: "Discard & Switch" or "Cancel"
- On confirm: reset dirty state, then switch
Pattern 3: Auto-Save with Visual Feedback
Use when: Simple inputs that should save immediately.
const [saved, setSaved] = useState(false);
const handleBlur = async () => {
if (value !== savedValue) {
await saveToBackend(value);
setSaved(true);
setTimeout(() => setSaved(false), 2000);
}
};
return (
<div className="flex items-center gap-2">
<Input value={value} onChange={...} onBlur={handleBlur} />
{saved && (
<span className="text-green-600 text-xs flex items-center gap-1">
<Check className="w-3.5 h-3.5" /> Saved
</span>
)}
</div>
);
When to use which:
| Scenario | Pattern |
|---|---|
| Complex multi-field editor | Pattern 2 (confirmation dialog) |
| Simple single input | Pattern 3 (auto-save + feedback) |
| List item selection | Pattern 1 (key-based remount) + Pattern 2 |
Quality Gates
Pre-Commit Sequence
# Main project
bun run format
bun run lint:fix
bun run validate
bun run validate:ci-parity
# UI project (if changed)
cd ui
bun run format
bun run lint:fix
bun run validate
Validate Runs
| Project | Command | Checks |
|---|---|---|
| Main | bun run validate |
typecheck + lint + format:check + test:fast |
| UI | bun run validate |
typecheck + lint + format:check |
Conventional Commits
All commits must follow conventional commit format:
<type>(<scope>): <description>
Types
| Type | When to Use | Version Bump |
|---|---|---|
feat |
New feature | MINOR |
fix |
Bug fix | PATCH |
perf |
Performance | PATCH |
docs |
Documentation | None |
style |
Formatting | None |
refactor |
Code restructure | None |
test |
Tests | None |
chore |
Maintenance | None |
Examples
# Correct
git commit -m "feat(cliproxy): add OAuth token refresh"
git commit -m "fix(doctor): handle missing config gracefully"
git commit -m "refactor(ui): split provider-editor into modules"
# Incorrect - REJECTED
git commit -m "added new feature"
git commit -m "Fixed bug"
Anti-Patterns to Avoid
1. God Files
// BAD: One file doing everything
// src/utils.ts (2000 lines with mixed concerns)
// GOOD: Split by domain
// src/utils/ui/colors.ts
// src/utils/ui/boxes.ts
// src/utils/shell-executor.ts
// src/utils/config-manager.ts
2. Barrel Import Bypass
// BAD: Direct import bypassing barrel
import { detectPlatform } from '../cliproxy/platform-detector';
// GOOD: Import from domain barrel
import { detectPlatform } from '../cliproxy';
3. Inline Everything
// BAD: Huge inline functions in components
function Component() {
const handleComplexOperation = () => {
// 100 lines of logic...
};
}
// GOOD: Extract to hooks or utilities
function Component() {
const { handleComplexOperation } = useComplexOperation();
}
4. Type Duplication
// BAD: Same types defined in multiple files
// file1.ts
interface Config { ... }
// file2.ts
interface Config { ... }
// GOOD: Single source of truth
// types/config.ts
export interface Config { ... }
5. Config Priority Pattern
When resolving configuration from multiple sources, follow this priority order:
// proxy-config-resolver.ts pattern
// Priority: CLI flags > Environment variables > config.yaml > defaults
const resolved = {
...DEFAULT_CONFIG, // 4. Defaults (lowest)
...yamlConfig, // 3. config.yaml
...envConfig, // 2. Environment variables
...cliFlags, // 1. CLI flags (highest)
};
This pattern is used in:
src/cliproxy/proxy-config-resolver.ts- Remote proxy configsrc/config/unified-config-loader.ts- Main config loading
Lint Enforcement Gates
Two ESLint gates (eslint.config.mjs) lock in the maintainability epic's gains:
ccs/no-new-throw-error(error): flags newthrow new Error(...). Use a typed error fromsrc/errors/error-types.ts(AuthError,ConfigError,ProfileError,ProviderError,NetworkError,ProxyError,MigrationError,ValidationError,RetryableError) sohandleErroremits a differentiated exit code. Existing ~340 sites are grandfathered ineslint-rules/throw-error-baseline.json; only new violations error. Regenerate the baseline when intentionally grandfathering a new site, or quarterly to prune converted entries:node scripts/generate-throw-error-baseline.jsmax-lines(warn, 400): warns on source files over 400 lines (skipBlankLines,skipComments). Split via the Monster File Splitting methodology above (barrelindex.tspreserves the public API).
When the no-throw rule blocks a change, prefer converting to the matching typed error. Only add to the baseline when the throw is genuinely out of scope to convert (and regenerate the baseline so the entry is explicit, not silent).
Related Documentation
- Codebase Summary - Full directory structure
- System Architecture - Architecture diagrams
- CLAUDE.md - AI-facing development guidance