Agent skills for shipping software

Design a structure, deepen an existing one, or harden it with guardrails.

Next.js turborepo with Blode UI, Ultracite (oxlint/shadcn), GitHub, and Vercel.

TypeScript CLI package: ESM, tsdown, vitest, changesets, CI.

Tenant isolation, routing, and custom domains on Cloudflare or Vercel.

What the interface should do, before anyone builds it.

Visual direction, design-system extraction, Tailwind builds, screenshot to markup, dark mode, responsive, and a React and Next.js UX audit with a ship verdict.

Runs nine browser probes, reviews before/after captures and suspicious duplicate states, proves the fix cleared, and runs an exploratory pass that tries a finished feature like a real user.

Builds and maintains a product's own verify CLI, doctor command, feature map, and worktree isolation, with reusable API checks, checked flow cleanup, and a reproduce-first bug handoff. Call by name.

Springs, gestures, easing, and curves pulled from a screen recording.

Decks with a story spine, speaker notes, and a contrast-checked QA pass, as Marp markdown, a web app, or a handoff to pptx.

Explains in the form that fits: plain prose, controlled English modelled on ASD-STE100, diagrams, interactive HTML pages, or narrated explainer videos, with exact technical terms and house vocabulary.

Writes anything as you or as your company: messages, posts, review comments, tickets, PRDs, design docs, talk scripts, copy, docs, READMEs, with the AI tells stripped. Works with no profile, writes one from pasted samples, and learns from your edits.

Agentic experience audit: 24 rules for tool parity, approval gates, and escape hatches, ship verdict.

Libraries, CLIs, SDKs, npm packages: 38 rules, agent-friendly checks, root-cause findings.

Implements AFDocs, Is Agentic, Is It Agent Ready, and agent 404 findings: llms.txt maps, markdown twins, API errors.

78 rules: punctuation, fonts, sizing, spacing, hierarchy, pairing.

SEO, AEO and GEO audits and fixes, search-demand research, writer briefs, and search, AI visibility, and conversion measurement.

Prunes low-value tests across a whole suite to a measured target with coverage held; invoked explicitly for a repo-wide or subsystem-wide campaign.

Writes and reviews executable plans with repository evidence, vertical slices, and explicit acceptance criteria.

Diff or PR review with file:line findings in confirmed and plausible tiers (bugs, structure, AI slop, security, new-test gating), report-only by default; apply mode lands the fixes and simplifies the diff.

PRs with short human descriptions, Linear IDs, templates, drafts, tidied commits.

Watches an open PR: conflicts, CI, comments. Fixes what it can.

npm releases with changesets: fix loop, CI watch, Version Packages merge, OIDC publish verify.

Finds where changes wait from push to deploy, then fixes the critical path, check placement, merge queue or post-merge revert, flaky tests, and deploy rate, and installs an agentic merge loop on GitHub (no merge queue, post-merge bisect and revert, agent re-lands and repairs), with a ledger.

Recover decisions, previous fixes, and context from past AI conversations with fast local search and source evidence.

Wires a repo so Claude Code, Codex and Cursor read the same instructions, migrates to native AGENTS.md without Claude wrappers, then audits and refactors instructions with grades and holds them to a size budget.

Creates and audits skills by their added value, with portable workflows, a validator, and regression scenarios.

Writes a named source as a markdown file the next turn can reread.