agentsclimarketplace

Code quality

Skill sordi-ai/skill-everything/skills/code-quality

Git-versioned agent memory: agents that never make the same mistake twice. Anthropic-Skill folder standard, multi-runtime (Claude Code, Cursor, Gemini CLI, OpenCode).

Install
npx -y skills add sordi-ai/skill-everything --skill code-quality

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 17 stars17 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Apply when writing or refactoring code. Generic rules to prevent the most common review comments — function length, naming, error handling, security, and tooling.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

5.9 KB, as published. Nobody here has run it

Sub-Skill: Code Quality & Common Mistakes

<!-- target: ~1800 tokens (real tiktoken count, see tools/render_readme_table.py) | 28 rules with severity classification -->

Purpose: Prevents the 20 % of mistakes that cause 80 % of review comments. Concrete rules from real projects — no boilerplate.

Rule classification

  • MUST — load-bearing. Violating causes bugs, security issues, or cross-team confusion. Never break.
  • SHOULD — default behavior. Deviation needs a documented reason in the code or PR.
  • AVOID — usually wrong; documented exception inline where needed.

Where these rules don't strictly apply: test fixtures, generated code (parser tables, codegen output, schema-derived types), CLI bootstrap scripts, framework adapters, and migration scripts may legitimately differ. The rules below apply to production code paths unless an inline exception says otherwise.


Rules

Functions & Logic

  1. SHOULD: Functions under ~30 lines. Longer → split where practical. Reduces cognitive load.
  2. AVOID: Boolean as last argument. render(true) is unreadable. Use enum or named-property: render({ withHeader: true }).
  3. SHOULD: Early return over nested if-blocks. Nesting depth > 2 is a warning sign.
  4. SHOULD: No magic numbers. if (status === 3)if (status === OrderStatus.SHIPPED).
  5. AVOID: Double negation. if (!isNotValid)if (isValid).

Variables & Naming

  1. SHOULD: Variable names describe content, not type. userList instead of arr, activeUserId instead of id.
  2. SHOULD: Temporary variables for complex expressions. const isEligible = age >= 18 && !isBanned; instead of cramming it all into an if.
  3. AVOID: Abbreviations except established ones (id, url, ctx, req, res). usr, cfg, tmp → spell out.

Error Handling

  1. MUST: Every async call needs try/catch or .catch(). Unhandled promise rejections crash Node processes.
  2. MUST: Never silently swallow error objects. catch (e) {} is forbidden. At minimum: logger.warn(e).
  3. MUST: Error messages must include context. throw new Error('User not found: ' + userId) not throw new Error('Not found').
  4. SHOULD: Catch specific exception types, not broad categories. Catch ConnectionError not Exception. Broad catches mask unrelated bugs and make debugging harder.
  5. SHOULD: Use exponential backoff with jitter for retry logic. Never retry in a tight loop. Each retry should wait longer than the last with randomization to prevent thundering herds.
  6. MUST: Never retry non-idempotent operations without explicit safeguards. A retried POST that creates a resource can produce duplicates. Use idempotency keys or check-before-retry patterns.
  7. SHOULD: Propagate errors with context rather than swallowing and re-raising generic exceptions. Use raise NewError("context") from original_error to preserve the chain. Reference: ERR-2026-016.

Imports & Dependencies

  1. SHOULD: No circular imports. When in doubt: run madge --circular src/.
  2. SHOULD: External dependencies only in dedicated adapter files. No direct axios.get() in business logic — always wrap through an interface.
  3. MUST: Remove console.log before commit. Use a pre-commit hook or linter rule no-console. Exception: debug helpers behind a build-time __DEV__ flag.

Tests

  1. MUST: Every new function needs at least one happy-path test. No merge without test coverage for new logic. Exception: pure projection wrappers, single-line getters, and deprecated shims may rely on integration tests.

Performance

  1. MUST: No DB queries in loops (N+1 problem). Instead of for (item of items) { await db.find(item.id) } → use a JOIN or WHERE id IN (...) query.
  2. MUST: Paginate large datasets. Never SELECT * FROM table without LIMIT. API endpoints with lists need ?page= and ?limit=.
  3. SHOULD: New WHERE clauses need indexes. Before every new query: is there a matching DB index? Without index → full table scan on growing data.
  4. SHOULD: Lazy loading for heavy operations. Load data only when needed, not "just in case".

Security

  1. MUST: Never use user input directly in queries. Always use prepared statements or ORM query builders. Applies to SQL, NoSQL, shell commands, and template engines.
  2. MUST: New API endpoints need auth. No endpoint without authentication and authorization — not even "internal" endpoints. Exception: documented public-asset / health-check routes that explicitly opt out in code comments.
  3. MUST: Never put secrets in code. No API keys, passwords, or tokens in source code or comments. Always use environment variables.
  4. MUST: Validate and sanitize user input. At the system boundary (API entry): check type, length, format. Never trust blindly.

Tooling & Cross-Platform

  1. MUST: Always prefix docker run -v from Git Bash / MSYS on Windows with MSYS_NO_PATHCONV=1. MSYS auto-converts unquoted Linux paths to Windows paths before passing them to docker.exe, mangling bind-mount arguments. Use MSYS_NO_PATHCONV=1 docker run -v "C:/path:/repo" ... with explicit Windows-style host paths. Reference: ERR-2026-013.

Why This Sub-Skill Earns Stars

Without these rules, the agent produces code that works but immediately gets flagged in reviews. With these rules, it writes code that looks like it came from a senior developer — on the first try. The MUST/SHOULD/AVOID classification means rules are followed strictly where they matter (security, error handling) and pragmatically where context matters (function length, lazy loading).

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.