跳到正文
cccskills
免费在 GitHub 公开

vibe-audit

Runs a comprehensive 20-point code quality audit on a "vibe-coded" codebase that was shipped fast and now needs hardening. Use this skill whenever the user mentions vibe coding, vibe audit, code audit, code review, code cleanup, hardening a prototype, productionizing a codebase, technical debt cleanup, or asks Claude to "audit my code", "review my codebase", "clean up my project", or check a freshly-shipped project for common quality issues. Also use when a user shares a repo and asks what's wrong with it or how to make it production-ready.

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(1)

  • SKILL.md4.3 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Vibe Audit

A 20-point audit for codebases that were shipped fast and need hardening. Work through every check in order. For each check: search the codebase, report findings with file paths and line numbers, then apply the fix (or propose it clearly if the change is large). Do not skip checks — report "none found" rather than omitting.

The 20 Checks

  1. Duplicate utilities. Search for duplicate utility functions across the codebase. Consolidate all duplicates into a single shared utility file and update every import.

  2. Secrets in config. Scan all committed config files for secrets or credentials. Flag anything that should have been an environment variable and move it.

  3. Giant functions. Find any functions over 400 lines. Break them into smaller, single-responsibility functions with clear, descriptive names.

  4. Giant components. Find any component over 200 lines. Split data fetching, business logic, and UI into separate layers.

  5. Dead code. Scan for functions and variables that are defined but never called or referenced. Remove all dead code and flag anything that's unclear.

  6. Silent catches. Find all empty or silent catch blocks. Add meaningful error logging and make sure failures are surfaced to the caller or the user.

  7. API call states. Find every API call in the UI. Verify each one has a loading state, an error state, and prevents duplicate requests on re-render.

  8. DB in route handlers. Check if any database queries are written directly inside route handlers. Extract all data access logic into a dedicated service or repository layer.

  9. Sync I/O in handlers. Search for synchronous I/O operations inside request handlers. Replace with async equivalents so the event loop stays unblocked.

  10. Unbounded lists. Check every endpoint that returns a list. Add pagination to any that fetch unbounded results.

  11. Inconsistent response shapes. Audit all API endpoints for consistent response shapes. Standardize to one envelope structure and update every endpoint and client that doesn't match.

  12. Floats for money. Search for any place floating point numbers are used for currency or financial values. Replace with integer arithmetic in cents and convert only at the display layer.

  13. String dates. Find all dates stored as plain strings. Convert to ISO 8601 format or Unix timestamps and ensure timezone context is always preserved.

  14. No retry on external calls. Find every external API call in the codebase. Add retry logic with exponential backoff to any that currently fail silently on a single error.

  15. Stale comments. Find all code comments in the codebase. Verify each one still accurately describes the code it sits next to and delete or rewrite anything that no longer matches.

  16. Unvalidated input. Find every place user input is accepted. Add validation and sanitization before it touches any business logic, database query, or API call.

  17. Missing auth. Check every API route for missing authentication middleware. Add auth checks to any endpoint that should require a logged-in user.

  18. Missing indexes. Check for missing indexes on columns used in WHERE clauses or joins. Add indexes to any high-frequency query columns.

  19. N+1 queries. Find any place related data is fetched inside a loop. Replace with a batched query or join.

  20. Duplicated SDK init. Search for third-party SDKs that are initialized in multiple places. Consolidate into a single shared instance.

How to Run

Go through the checks in order. For each: state the check, show what you searched, list findings (path:line), then fix or propose the fix. At the end, give a summary table of all 20 checks with status (fixed / proposed / none found). Ask the user before making sweeping changes that touch many files.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

adhd

免费

Formats output for ADHD-friendly action — lead with the next step, number tasks, restate state, cut fluff. Use when named (adhd), when the user mentions ADHD or focus/executive-function constraints, or when they ask for steps they can execute immediately.

中文简介正在准备中。当前显示原文说明。

nateslabach/skills62026年9月29日 更新

Devil's-advocate attack on AI workflow designs — threat model, assumption challenges, code critiques, and hypothetical redesigns with no file edits. Use for "attack this design", devil's advocate, or hostile challenge on agents/tools/summarization/tokens. Not for PR peer review (code-review) or self-review that applies fixes (ai-systems-review). Prefer: this skill → decide → ai-systems-review.

中文简介正在准备中。当前显示原文说明。

nateslabach/skills62026年9月29日 更新

Hostile self-review of recently written AI workflow code that then applies fixes (agents, tools, chat context, summarization, tokens, attachments). Use after generating AI systems code, or when asked to self-review and harden it. Not for PR peer review (code-review) or no-edit devil's-advocate attacks (adversarial-challenge). Prefer: adversarial-challenge → decide → this skill.

中文简介正在准备中。当前显示原文说明。

nateslabach/skills62026年9月29日 更新

Writes in ASD-STE100 simplified technical English — short sentences, plain words, active voice. Use when the user wants terse technical prose, names this skill, or asks for STE100.

中文简介正在准备中。当前显示原文说明。

nateslabach/skills62026年9月29日 更新

branding

免费

Builds personal brand assets using Seth Godin's frameworks: signature symbol, slogan, surprise (Purple Cow), salient idea, and signature story. Use when named (branding), or when the user asks for a brand symbol, slogan, tagline, Purple Cow, counterintuitive angle, core idea, brand story, or personal brand positioning.

中文简介正在准备中。当前显示原文说明。

nateslabach/skills62026年9月29日 更新

Disables the AI's default hedging, diplomatic softening, and social-nicety layers to deliver cold, unfiltered, consequentialist analysis on any topic. Use when the user wants the AI's actual assessment without padding, regardless of how harsh or uncomfortable the honest answer is.

中文简介正在准备中。当前显示原文说明。

nateslabach/skills62026年9月29日 更新

查看 nateslabach 的全部技能

报告此技能的问题