Code Consistency Skill

Enforces, detects, reviews, and generates consistent code across all registered languages. Includes inline quality triage — lightweight checks that flag issues for deeper analysis via the code-quality skill when warranted.

Language Registry

Each supported language has a reference file in references/. To add a new language, create references/<lang>.md following the four-section template (see below).

Adding a New Language

When encountering a language not in the registry:

1. Create references/<lang>.md with the four standard sections:
   • Naming Conventions
   • Error Handling
   • Project Structure & Packaging
   • Logging / Observability
2. Add a row to the registry table above
3. Update the skill description's language list
4. Use authoritative style guides as baseline (official lang docs, widely adopted community guides)

Core Philosophy

• Detect first, enforce second. Never impose external style onto a codebase with its own conventions. Scan what's there and match it.
• Pragmatic strictness. Flag critical issues (bugs, ambiguity, maintenance traps). Suggest minor issues. Never block progress over cosmetics.
• Quality triage. Surface obvious performance, security, testability, and architecture smells inline. Recommend the code-quality skill for deep analysis.
• Context-driven output. Match format to task:
  • Inline comments → editing a specific file
  • Numbered violation report → broad review
  • Before/after diff → rewriting or refactoring
  • Style guide doc → reference material request

Workflow

Step 1 — Detect the Language(s)

Identify which language(s) the task involves. Check the registry. If the language isn't registered and the task is substantial, offer to create a reference file.

Step 2 — Load Reference Files

Before writing any code or review output, read the appropriate reference(s):

• Single language → read references/<lang>.md
• Multi-language (e.g., Terraform + Bash) → read both
• IaC + config tasks → check if Ansible or Terraform references apply

Step 3 — Style Detection (when existing code is present)

Scan provided code to infer project conventions. Use the Style Detection Checklist below. Record findings as a brief internal "style snapshot" before proceeding.

Step 4 — Execute the Task

Step 5 — Quality Triage (inline)

While executing, flag any of these if spotted:

• 🔴 Performance: O(n²)+ in hot paths, unbounded allocations, blocking I/O in async contexts, missing index hints
• 🔴 Security: hardcoded secrets, injection vectors, path traversal, missing input validation, permissive CORS/auth
• 🟡 Testability: god functions (>50 LOC with multiple concerns), hidden dependencies, global state, untestable side effects
• 🟡 Architecture: circular imports, abstraction leaks, SRP violations, inappropriate coupling between layers

If more than 2 quality flags fire, recommend: "Consider running the code-quality skill for deep analysis on [performance|security|testability|architecture]."

Step 6 — Present Output

Use format appropriate to context. Always separate critical from suggested.

Style Detection Checklist

Run this mentally when existing code is provided.

Naming

• Case convention per symbol type (snake_case, camelCase, PascalCase, kebab-case)
• Prefix/suffix patterns (e.g., _internal, I-prefix, Err-prefix, $-prefix)
• Constant casing
• Test naming patterns

Error Handling

• Errors wrapped with context? What mechanism?
• Sentinel/custom error types?
• Boundary behavior (panics, exceptions, exit codes, trap handlers)

Project Structure

• Flat vs nested layout
• Interface/trait placement
• Module system conventions
• Config file conventions

Logging/Observability

• Which logging library/mechanism?
• Structured vs unstructured?
• Log level conventions
• Trace/span propagation?

IaC-Specific (Terraform/Ansible)

• Resource naming conventions
• Module/role structure
• Variable/parameter organization
• State/inventory management patterns

Severity Levels

Reference Files

Read before working on language-specific tasks:

• references/python.md — PEP 8 naming, error handling, packaging, logging, OpenTelemetry
• references/golang.md — Effective Go + Uber naming, error handling, module layout, logging, OpenTelemetry
• references/bash.md — Shell naming, error handling (set -euo, traps), script structure, logging
• references/rust.md — RFC 430 naming, Result/Error handling, Cargo layout, tracing crate
• references/c.md — Linux kernel style naming, return-code error handling, Makefile/CMake layout, syslog + custom macros
• references/powershell.md — Verb-Noun naming, error handling, module structure, logging
• references/terraform.md — HCL naming, error handling, module layout, state management
• references/ansible.md — YAML conventions, error handling, role structure, logging/verbosity