Technical spec
Skill shinpr/ai-coding-project-boilerplate/.claude/skills-en/technical-spec
Agentic coding TypeScript boilerplate for Claude Code: sub-agent workflows with built-in quality checks and context engineering.
npx -y skills add shinpr/ai-coding-project-boilerplate --skill technical-specAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
What its author says it does
Copied from the file, not written here
Defines environment variables, architecture design, and build/test commands. Use when configuring environment or designing architecture.
SKILL.md
6.4 KB, as published. Nobody here has run it
Technical Design Rules
Prerequisite Detection
Inspect manifests, lockfiles, build/test configuration, CI definitions, and representative source files before applying a technology- or command-specific rule. Treat a tool, script, path alias, or runtime as observed only when repository evidence names it. Label conclusions from surrounding patterns as inferred. When a missing decision changes architecture, compatibility, security, or verification, stop and name the exact configuration or user decision required.
Basic Technology Stack Policy
These rules apply to a TypeScript application when repository configuration confirms that stack. Select architecture by mapping current requirements and accepted constraints to explicit module responsibilities, dependency direction, data flow, and verification boundaries.
Environment Variable Management and Security
Environment Variable Management
- Centrally manage environment variables and build mechanisms to ensure type safety
- Read environment variables through one typed configuration boundary; application code consumes validated configuration values
- Give a variable a default only when requirements define valid behavior for absence; otherwise fail configuration validation with the variable name and expected format
Security
- Keep local
.envfiles outside version control and provide non-secret example files for required variable names - Load API keys and secrets from the configured secret store or runtime environment boundary
- Log and return only fields approved for the current trust boundary; redact credentials, tokens, personal data, and internal diagnostics from untrusted responses
Architecture Design
Architecture Design Principles
Select architecture using these observable decisions:
- Responsibilities: Each module/layer names the behavior it owns and the behavior it delegates
- Dependency direction: Imports and runtime calls follow the project boundary rules observed in configuration or representative implementations
- State/data ownership: Each persisted or mutable value has one authoritative owner
- Verification boundary: Each public contract has a unit, integration, or E2E check that can observe it
Unified Data Flow Principles
Basic Principles
- Single Data Source: Store the same information in only one place
- Structured Data Priority: Use parsed objects rather than JSON strings
- Responsibility Separation: Each layer names the data or behavior it owns and the boundary through which other layers use it
Data Flow Best Practices
- Validation at Input: Validate data at input layer and pass internally in type-safe form
- Centralized Transformation: Consolidate data transformation logic in dedicated utilities
- Structured Logging: Output structured logs at each stage of data flow
Build and Testing
Select the package manager from the packageManager field, lockfile, or established CI command in that order. Execute only scripts present in the selected manifest.
Build Commands
build- TypeScript buildtype-check- Type check (no emit)
Testing Commands
test- Run teststest:coverage- Run tests with coveragetest:coverage:fresh- Run tests with coverage (fresh cache)test:safe- Safe test execution (with auto cleanup)cleanup:processes- Cleanup Vitest processes
Quality Assurance Mechanism Awareness
Before executing quality checks, identify what quality mechanisms exist for the change area:
- Primary detection: inspect the change area's file types, project manifest, and configuration to identify applicable quality tools
- Check CI pipeline definitions for checks that cover the affected paths
- Check for domain-specific linter or validator configurations (e.g., schema validators, API spec validators, configuration file linters)
- Check for domain-specific constraints in project configuration (naming rules, length limits, format requirements)
- Supplementary hint: IF task file specifies Quality Assurance Mechanisms → use them as additional hints for which domain-specific checks to look for
- Include discovered domain-specific checks alongside standard quality phases below
Quality Check Requirements
Quality checks are mandatory upon implementation completion:
Phase 1-3: Code Quality Checks
- Auto-detect and execute the following from package.json scripts:
- lint + format check
- Detect unused exports
- Detect circular dependencies
- TypeScript build
Transition evidence: every applicable static/domain check exits successfully. A missing required script is reported with the manifest/configuration path and blocks the next phase until an equivalent established command is identified.
Phase 4: Tests
test- Test execution
Transition evidence: all applicable configured test suites pass, or an environment-dependent suite is recorded as blocked with its exact prerequisite.
Phase 5: Code Quality Re-verification
check:code- Re-verify code quality (clean up side effects from test fixes in Phase 4)
Completion evidence: static/domain checks still pass after test-related fixes, the build succeeds, and every required test has passed or is explicitly blocked.
Auxiliary Commands
check:all- Overall integrated check (check:code + test) *for manual batch verificationopen coverage/index.html- Check coverage reportformat- Format fixeslint:fix- Lint fixes
Troubleshooting
- Port in use error: Run the
cleanup:processesscript - Cache issues: Run the
test:coverage:freshscript - Dependency errors: First record the failing resolver output, selected package manager, manifest, and lockfile state. Use the repository's established clean-install command only when it preserves the lockfile and generated artifacts; request approval before an operation that removes or regenerates dependency state
Coverage
- Treat coverage as a diagnostic signal for finding untested areas, not a target (a target gets gamed into trivial tests — Goodhart's Law). Concentrate tests on critical paths and business logic whose regression would matter
- Any enforced numeric threshold is the project's CI/coverage config, not a goal in itself
- Metrics (what coverage reports break down): Statements, Branches, Functions, Lines