mohamedcody/awesome-agent-skills

Read-only WCAG audit of UI and markup with a concrete fix per finding. Use when asked to check accessibility, a11y, keyboard or screen-reader support, or contrast.

Mines a codebase's real commands, conventions and boundaries into an agent-agnostic AGENTS.md, nested per package in monorepos, every rule backed by counted evidence. Use when asked to write, generate or refresh AGENTS.md or agent instructions for a repo.

Designs or reviews an HTTP API's shape before code: resources, versioning, pagination, idempotency, filtering. Use when asked to design an API or an endpoint.

Read-only audit of a whole codebase's design: module boundaries, dependency direction, docs-vs-code drift, extensibility, with a verdict. Use when asked to audit the architecture.

Fixes a bug through a runnable reproduction and its root cause. Use for errors, crashes, failing tests or builds, flaky behavior, or repeated failed fixes.

Checks public claims (site, README, store listing, privacy policy, structured data) against the code and config that decide them, then fixes drift. Use when asked whether the site still matches the product.

Behavior-preserving cleanup of AI-like code noise: narrating comments, dead code, vague names, over-abstraction. Use when asked to clean up code or comments, de-slop, or apply a slop-audit report.

Generates a full, cited technical documentation set for a codebase, local or from any git URL: architecture, module and API reference, data model, flows, config, operations, how-tos. Use when asked to document a repo.

Explains how a codebase works, local or from any git URL, as one illustrated page: purpose, stack, structure, entry points, main flow, where to start reading, every claim cited. Use when asked to explain a repo.

Reviews a diff, patch or pull request for correctness, security and team standards, with severity-ranked findings. Use when asked to review a change or PR, or before merging.

Handles review comments others left on the user's change: which points to apply, clarify or push back on, each checked against the code. Use when asked to address reviewer or bot comments.

Multi-model panel review of a code change: blind reviewers, then one cross-examined verdict. Use when asked for a second opinion, a multi-model or panel review, or a high-stakes review.

Coding conventions that fill the gaps a project's own rules leave: naming, structure, immutability, types, error handling, layering. Use when starting a module, setting lint rules, refactoring to conventions, or onboarding.

Plans a scheduled batch of platform-fit posts about a product from repos, sites or files, every claim traced to evidence. Use when asked for a content plan or product posts.

Adapts one finished image into a horizontal 16:9 and a vertical 9:16 version for social platforms without redrawing the artwork. Use when asked to resize a picture for posts.

Publishes prepared posts to the user's own accounts through their live browser, on schedule, with login checks, a no-duplicates ledger and a confirmation gate before anything goes public. Use when asked to publish posts.

Turns one source (article, guide, notes) into three posts for all platforms: a long read, a regular and a short post, in the language the user picks. Use when asked to adapt one text for social posts.

Builds a reusable profile of the user's own writing voice from their posts, files, samples or an interview, for other skills to read. Use when asked to learn the user's voice.

Writes short product copy: headlines, titles, meta and store descriptions, buttons, empty states, error messages, subject lines, CTAs. Use when asked for such copy.

Read-only audit of a database layer: schema anti-patterns, query and index fit, integrity, concurrency, migrations, tenancy, with a verdict. Use when asked to review a schema, migrations or data model.

Read-only supply-chain audit of packages and agent extensions: lockfiles, typosquats, dependency confusion, install scripts, licenses, CVEs. Use when asked if a package, manifest change or bot bump is safe.

Plans and executes dependency upgrades: batches by risk, changelog-driven majors, CVE overrides, one revertable commit per batch. Use when asked to upgrade packages, merge bot update PRs, or act on an audit report.

Writes a design doc or ADR: requirements and scale first, real alternatives with trade-offs, a grounded recommendation, non-goals, migration path; also runs a pre-mortem on a plan. Use when choosing an approach.

Line-edits Markdown documents into clear, specific, publication-ready prose: cuts bloat and template structure, fixes vague wording and number style. Use when asked to edit, tighten or polish a document.

Applies consistent error handling: typed errors, operational vs programmer errors, API error envelopes, HTTP status mapping, user-facing messages. Use when adding or refactoring error handling or designing an error contract.

Rewrites the author and committer on commits with a wrong identity, then force-pushes after access checks, a mirror backup and a confirmation gate. Use when a commit has the wrong name or email.

Plans a codebase as a bisectable series of commits that each build and test alone, delivered as one plan file. Use when asked to split a project into commits.

Erases a repository's history and rebuilds it as a curated commit series over the same tree, with a backup, confirmation gates and a tree-hash proof. Use when asked to rewrite history as readable commits.

Erases a repository's entire history into one fresh commit and force-pushes, after access checks, a mirror backup, a secret scan and a confirmation gate. Use when asked to wipe history.

Read-only recovery of every commit a repository ever held, force-pushed history included, from refs, PR refs, backups and the host's activity log. Use when asked what a repo held before a rewrite.

Proofreads prose and reports grammar, logic and flow issues as suggestions without rewriting the text. Use when asked to proofread or copy-edit while keeping the author's wording.

Removes signs of AI writing from English text: cliches, filler, jargon, sycophancy, em-dash habits, fake sources, chatbot markers like ':contentReference'. Use when text reads like a chatbot.

Read-only audit of a landing page's conversion structure: primary CTA, form friction, message match, trust elements, above-the-fold value. Use when a page is not converting.

Audits a public client (extension, app, SPA, CLI, SDK) for what it reveals about the private backend (limits, anti-abuse, backdoors, infra) and client hardening. Use before open-sourcing or a store submission.

Applies structured logging: levels, PII redaction, request logging, wide events (canonical log lines). Use when adding or reviewing logs, choosing levels, or after logs failed or leaked data in an incident.

Runs an authorized penetration test (web app, API, network, cloud, mobile) per PTES and OWASP, with CWE/CVSS findings and a retest. Use for a pentest or red team engagement against a live target.

Read-only performance and reliability audit: event loop, backpressure, memory and CPU, timeouts, retries, queues, Core Web Vitals, bundles, with a verdict. Use when something is slow, memory climbs or latency spikes.

Simplifies a written plan, design or spec against the system it changes: over-engineering, redundant guards and contradictions, as ranked cuts in a separate document. Use when asked what a plan can drop.

Read-only privacy audit of a codebase: what personal data it holds, where it flows, minimisation, retention and deletion, data-subject rights, consent gating, exposure, with a verdict. Use for a privacy or GDPR/CCPA readiness review, before launch in a new region, or after adding a tracker or SDK.

Read-only sweep for regressions against a recorded baseline: typecheck, lint, test suites, wire contracts, generated-file drift, downstream builds, deploy drift. Use to verify everything before or after a deploy.

Root-cause analysis for incidents, recurring failures and process problems with no failing test: 5 Whys, fishbone, PDCA, A3. Use for a postmortem or why something keeps happening.

Read-only security audit of source code: injection, secrets, authn and authz, dependency CVEs, CI/CD exposure, crypto misuse, with CWE-mapped fixes. Use for a security review, before a release, or after adding auth.

Read-only SEO audit of a site or web codebase: crawling, indexing, on-page, structured data, rendering, hreflang, links, visibility to LLMs and agents, baseline diffs. Use when asked to audit SEO, titles or meta.

Removes installed agent skills from every AI agent on the machine (Claude Code, Codex, Gemini CLI, Cursor...) behind a keep list, an archive and a confirmation gate. Use when asked to uninstall installed skills.

Read-only audit of a repository for AI-slop markers in code, comments, tests, docs and CI, each verified against the code and ranked. Use when asked to find AI slop or machine-written code.

Learns a website's writing style by crawling it in a live browser, writes a reusable style guide, then rewrites documents in that voice. Use when asked to learn a site's style or write in its voice.

Read-only audit of which tests are missing, weak, stale or mis-scoped, mapped by what tests actually call, ranked P0-P3 by risk, with a verdict and a worklist for awesome-test-writing. Use when asked what tests are missing, where the coverage gaps are, or to audit a test suite.

Designs and writes tests that catch real regressions: unit, integration or E2E placement, fixtures, behavior-first assertions, characterization and property tests. Use when asked to write tests, add coverage or a regression test.

Translates Russian text of any genre into natural English, keeping structure, formatting, links, identifiers and the author's voice; can mirror a folder. Use when asked to translate Russian into English.

Review UI code for Web Interface Guidelines compliance. Use when asked to "review my UI", "check accessibility", "audit design", "review UX", or "check my site against best practices".