59 public agent skills from 14 publishers, placed on the delivery lifecycle. Every one is somebody else's — install it from the registry, not from here.
Redline reviews the diff. It has no opinion on how the diff gets written — but the engineers asking for the standard keep asking the next question, which is what to give the agent so the diff arrives in better shape. A skill is the answer: a file the model loads when it recognises the situation, carrying procedure it would otherwise guess at. It is nearly free — it loads, it shifts the output, you move on.
Something that takes the wheel instead — spawns subagents, drives a browser, keeps going on its own — is on the agents page (24 of them). That split is a judgement we made, not a field the registry publishes: skills.sh lists both as skills, and we separated them because the question you ask before installing one is not the question you ask before installing the other. Everything below is knowledge applied to work you are already doing.
Ordered by delivery phase. Anything that spans two phases is listed under both — the phase chips on each card say which.
Turning an intention into something an agent can execute without inventing the requirements as it goes.
Pulls a vague idea apart into options and trade-offs before anything is committed to a plan.
Reach for it when The ticket says what somebody wants but not what should be built, and the first design you thought of is the only one you have thought of.
npx skills add obra/superpowersInstalls all of superpowers, not this one alone.
The plan format the rest of the superpowers library executes against — ordered, checkable steps rather than prose.
Reach for it when Before any change big enough that you would not want the agent deciding the order of the work on its own.
npx skills add obra/superpowersInstalls all of superpowers, not this one alone.
Write the specification first, then let the implementation be judged against it rather than against taste.
Reach for it when Work that will be reviewed by someone who was not in the room when it was scoped.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Decomposes a feature into tasks small enough that each one can be finished and verified independently.
Reach for it when A feature that is obviously more than one sitting, and you want the pieces to land separately.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
States the constraints — budget, latency, compatibility, blast radius — as inputs rather than discovering them in review.
Reach for it when The requirement that will actually kill the design is non-functional, and nobody has written it down.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Deliberately curating what the agent sees and when — the single biggest lever on output quality.
Reach for it when Output quality is sliding: hallucinated APIs, ignored conventions, patterns from the wrong part of the repo.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Shapes intent into a measurable objective with explicit evidence and bounded scope, rather than a description of activity.
Reach for it when The task is phrased as work to do ('improve the dashboard') instead of an outcome you could check.
npx skills add openai/skillsInstalls all of skills, not this one alone.
Makes the agent stop and ask instead of filling a gap in the requirements with a plausible guess.
Reach for it when Any task where a confidently wrong assumption costs more than the round trip of asking.
npx skills add trailofbits/skillsInstalls all of skills, not this one alone.
Interfaces, module boundaries and domain language — decided before the code sets them in concrete.
Designing interfaces that are hard to misuse and cheap to evolve — naming, versioning, error shape, defaults.
Reach for it when You are about to add a public function, endpoint or package export that other teams will depend on.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Where module boundaries go, and what belongs on each side of one.
Reach for it when The new feature does not have an obvious home, and the wrong answer becomes permanent within a sprint.
npx skills add mattpocock/skillsInstalls all of skills, not this one alone.
Modelling the problem so that illegal states cannot be represented, instead of validating them at every call site.
Reach for it when You are reaching for a fourth optional boolean on the same type.
npx skills add mattpocock/skillsInstalls all of skills, not this one alone.
One name per concept, shared between the code and the people who asked for it.
Reach for it when The same thing is a `customer` in one service, an `account` in the next and a `user` in the database.
npx skills add mattpocock/skillsInstalls all of skills, not this one alone.
Anthropic's guidance for producing interfaces that look designed rather than defaulted.
Reach for it when The agent is generating UI and everything comes out looking like an untouched component library.
npx skills add anthropics/skillsInstalls all of skills, not this one alone.
Vercel's design rules for agent-generated web interfaces: layout, spacing, typography, motion, state.
Reach for it when You want generated UI to be consistent across a team without a designer reviewing every PR.
npx skills add vercel-labs/agent-skillsInstalls all of agent-skills, not this one alone.
Produces accurate architecture and data-flow diagrams from the code that exists, not the code somebody described.
Reach for it when Onboarding onto an unfamiliar service, or preparing a design review for a system nobody has drawn in two years.
npx skills add trailofbits/skillsInstalls all of skills, not this one alone.
Finds the places where deepening a module would remove a whole class of call-site complexity.
Reach for it when Every feature touches nine files and you want to know which boundary is wrong.
npx skills add mattpocock/skillsInstalls all of skills, not this one alone.
Recording architectural decisions so the reasoning survives the people who made it.
Reach for it when A decision that will look arbitrary in a year unless the alternatives are written down now.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Writing the change: framework-specific conventions, and the discipline that keeps a long agent run on the rails.
Anthropic's guidance for producing interfaces that look designed rather than defaulted.
Reach for it when The agent is generating UI and everything comes out looking like an untouched component library.
npx skills add anthropics/skillsInstalls all of skills, not this one alone.
Keeps the tree working after every step, so a failure is attributable to the last change rather than to the batch.
Reach for it when Any refactor big enough that a single broken commit would be painful to bisect.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Component structure, state placement, accessibility and the rendering costs that only show up on a real device.
Reach for it when Building UI that has to survive contact with slow networks and assistive technology.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Gives each agent run its own checkout, so parallel work does not fight over one working directory.
Reach for it when You want to run more than one agent on the same repository without them stepping on each other.
npx skills add obra/superpowersInstalls all of superpowers, not this one alone.
Vercel's current React guidance — server and client boundaries, data fetching, caching, the Compiler era.
Reach for it when Your agent is still writing 2022 React: manual memoisation, effects for derived state, client components by default.
npx skills add vercel-labs/agent-skillsInstalls all of agent-skills, not this one alone.
React Native conventions and the places the web mental model breaks.
Reach for it when A web-shaped agent is writing mobile code and the result runs badly on device.
npx skills add vercel-labs/agent-skillsInstalls all of agent-skills, not this one alone.
File-based routing for Expo apps, from Expo — layouts, groups, typed routes, deep links.
Reach for it when Starting or restructuring navigation in an Expo app, where the generated code otherwise lags the current API by a major version.
npx skills add expo/skillsInstalls all of skills, not this one alone.
Cloudflare's own guidance for Workers: runtime limits, bindings, request lifecycle, what does not exist at the edge.
Reach for it when Writing Workers code, where half the Node.js API surface an agent assumes is simply absent.
npx skills add cloudflare/skillsInstalls all of skills, not this one alone.
Schema, indexing and row-level security as Supabase recommends them.
Reach for it when Anything touching RLS policies — the place where a plausible-looking generated policy is a data leak.
npx skills add supabase/agent-skillsInstalls all of agent-skills, not this one alone.
The current Prisma Client surface, from Prisma.
Reach for it when Generated Prisma queries keep using APIs that moved or were removed a major version ago.
npx skills add prisma/skillsInstalls all of skills, not this one alone.
Building ASP.NET Core web APIs the way the .NET team builds them.
Reach for it when Starting a service, or bringing a controller-era codebase onto the current minimal-API conventions.
npx skills add dotnet/skillsInstalls all of skills, not this one alone.
The strict red-green-refactor loop, written so an agent cannot quietly write the test after the code.
Reach for it when You want the test to be evidence the code works, not a transcript of what it already does.
npx skills add obra/superpowersInstalls all of superpowers, not this one alone.
WCAG in practice — semantics, focus management, contrast, and what automated checks cannot see.
Reach for it when Any user-facing UI. Generated markup passes axe and still cannot be operated with a keyboard.
npx skills add addyosmani/web-quality-skillsInstalls all of web-quality-skills, not this one alone.
Proving the change does what it claims — and finding the assertions that were never written.
The strict red-green-refactor loop, written so an agent cannot quietly write the test after the code.
Reach for it when You want the test to be evidence the code works, not a transcript of what it already does.
npx skills add obra/superpowersInstalls all of superpowers, not this one alone.
The same discipline in the SDLC set, so it composes with the planning and review skills beside it.
Reach for it when You are installing the Osmani set anyway and want one consistent vocabulary across the phases.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Forces the agent to actually run the thing before it claims to be finished.
Reach for it when The failure mode you keep hitting is a confident summary of work that does not build.
npx skills add obra/superpowersInstalls all of superpowers, not this one alone.
Authoring Playwright tests, shipped in the Playwright repository itself — locators, waiting, fixtures, traces.
Reach for it when Your generated E2E tests are full of arbitrary sleeps and brittle CSS selectors.
npx skills add microsoft/playwrightInstalls all of playwright, not this one alone.
Tests the invariant across generated inputs instead of the three examples you thought of.
Reach for it when Parsers, serialisers, money arithmetic, permission logic — anywhere the interesting input is the one you did not imagine.
npx skills add trailofbits/skillsInstalls all of skills, not this one alone.
Breaks the code on purpose to find out whether the suite notices — the only honest measure of test quality.
Reach for it when Coverage is high, confidence is not, and you suspect the tests assert very little.
npx skills add trailofbits/skillsInstalls all of skills, not this one alone.
Reading a diff critically, and the far harder skill of acting on what a reviewer said.
How to act on review feedback without either capitulating to all of it or arguing with all of it.
Reach for it when The agent's instinct on a review comment is to apply it immediately, including when it is wrong.
npx skills add obra/superpowersInstalls all of superpowers, not this one alone.
What to look for in a diff, in priority order, and what to leave alone.
Reach for it when Your AI reviewer produces forty style nits and misses the race condition.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Reduces complexity while preserving exact behaviour — judged on whether a new joiner reads it faster, not on line count.
Reach for it when The feature works and the tests pass, but the implementation is heavier than the problem.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Finds the places where deepening a module would remove a whole class of call-site complexity.
Reach for it when Every feature touches nine files and you want to know which boundary is wrong.
npx skills add mattpocock/skillsInstalls all of skills, not this one alone.
Retiring an old path safely: migrating consumers, proving nothing still calls it, then deleting it.
Reach for it when The new implementation shipped six months ago and the old one is still there because nobody dared.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Threat modelling, scanning, supply chain, and finding the other five copies of the bug you just found.
Schema, indexing and row-level security as Supabase recommends them.
Reach for it when Anything touching RLS policies — the place where a plausible-looking generated policy is a data leak.
npx skills add supabase/agent-skillsInstalls all of agent-skills, not this one alone.
Input validation, authn/authz, secrets, dependency risk and logging — the everyday floor rather than a formal audit.
Reach for it when The baseline you want loaded on any change that touches user input or credentials.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
OpenAI's secure-coding guidance for agent-written code.
Reach for it when A second, independently written baseline when you do not want one author's blind spots.
npx skills add openai/skillsInstalls all of skills, not this one alone.
Running Semgrep properly and writing rules that catch a pattern without drowning the repo in findings.
Reach for it when You want a class of bug caught deterministically in CI instead of hopefully by a model.
npx skills add trailofbits/skillsInstalls all of skills, not this one alone.
Writing and running CodeQL queries for dataflow problems a pattern matcher cannot see.
Reach for it when Taint tracking — proving that untrusted input can or cannot reach a dangerous sink.
npx skills add trailofbits/skillsInstalls all of skills, not this one alone.
Hardening CI: pinned actions, scoped tokens, and the injection paths in a pull_request_target trigger.
Reach for it when Any workflow that runs against a fork's code or holds a token worth stealing.
npx skills add trailofbits/skillsInstalls all of skills, not this one alone.
Branches, pipelines, releases and the deploy itself.
Hardening CI: pinned actions, scoped tokens, and the injection paths in a pull_request_target trigger.
Reach for it when Any workflow that runs against a fork's code or holds a token worth stealing.
npx skills add trailofbits/skillsInstalls all of skills, not this one alone.
The end-of-branch checklist: rebase, squash, description, and what must be true before it merges.
Reach for it when Agent-authored branches that arrive with 40 commits called 'fix' and an empty description.
npx skills add obra/superpowersInstalls all of superpowers, not this one alone.
Commit hygiene, branching and semantic versioning as one consistent set.
Reach for it when Commit messages that describe the diff instead of the reason for it.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Pipeline structure, caching, matrix builds and gating.
Reach for it when A pipeline that takes eleven minutes to tell you about a lint error.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Release readiness: rollout, flags, rollback and what gets watched after the deploy.
Reach for it when A change that is risky enough to want a plan for turning it off again.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Writing Actions workflows, from a team that maintains some very large ones.
Reach for it when Beyond the generated starter workflow — reusable workflows, concurrency, permissions.
npx skills add dotnet/skillsInstalls all of skills, not this one alone.
Expo Application Services build, submit and update pipelines, from Expo.
Reach for it when Automating mobile releases, where the store submission steps are the part that is easy to get wrong.
npx skills add expo/skillsInstalls all of skills, not this one alone.
Production reality: debugging, performance, instrumentation, and writing down why the system is the way it is.
Produces accurate architecture and data-flow diagrams from the code that exists, not the code somebody described.
Reach for it when Onboarding onto an unfamiliar service, or preparing a design review for a system nobody has drawn in two years.
npx skills add trailofbits/skillsInstalls all of skills, not this one alone.
Retiring an old path safely: migrating consumers, proving nothing still calls it, then deleting it.
Reach for it when The new implementation shipped six months ago and the old one is still there because nobody dared.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Reproduce, narrow, hypothesise, test one variable at a time — instead of changing four things and rerunning.
Reach for it when The bug has survived two speculative fixes and is starting to cost a day.
npx skills add obra/superpowersInstalls all of superpowers, not this one alone.
Reading failures properly, and getting an agent out of a loop where it keeps retrying the same broken approach.
Reach for it when The agent has been stuck on the same error for six turns and is getting more confident, not less.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Measure, find the actual bottleneck, change one thing, measure again.
Reach for it when Somebody says it feels slow and the instinct is to start adding caches.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
LCP, INP and CLS: what moves each one, and what only looks like it does.
Reach for it when A failing field-data report, where lab numbers are already green and unhelpful.
npx skills add addyosmani/web-quality-skillsInstalls all of web-quality-skills, not this one alone.
WCAG in practice — semantics, focus management, contrast, and what automated checks cannot see.
Reach for it when Any user-facing UI. Generated markup passes axe and still cannot be operated with a keyboard.
npx skills add addyosmani/web-quality-skillsInstalls all of web-quality-skills, not this one alone.
Cloudflare's view of web performance — caching, edge behaviour and delivery.
Reach for it when The bottleneck is in front of your origin rather than inside your bundle.
npx skills add cloudflare/skillsInstalls all of skills, not this one alone.
Finding and fixing what is slow or expensive in a Vercel-hosted app.
Reach for it when A bill or a p95 that moved and nobody knows which deploy did it.
npx skills add vercel-labs/agent-skillsInstalls all of agent-skills, not this one alone.
Traces, dumps and allocation analysis, from the team that builds the runtime.
Reach for it when A .NET service with a memory profile or GC pattern that needs real tooling, not guesses.
npx skills add dotnet/skillsInstalls all of skills, not this one alone.
What to emit — logs, metrics, traces — and what must never end up in any of them.
Reach for it when Adding instrumentation to a path that handles personal data, where a helpful log line is a privacy incident.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Recording architectural decisions so the reasoning survives the people who made it.
Reach for it when A decision that will look arbitrary in a year unless the alternatives are written down now.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
Finding skills, writing your own, and keeping a skill library from turning into a junk drawer.
Deliberately curating what the agent sees and when — the single biggest lever on output quality.
Reach for it when Output quality is sliding: hallucinated APIs, ignored conventions, patterns from the wrong part of the repo.
npx skills add addyosmani/agent-skillsInstalls all of agent-skills, not this one alone.
The index for the rest of the library — which skill applies when, so the others actually get invoked.
Reach for it when Install this alongside the superpowers set or most of it will sit unused.
npx skills add obra/superpowersInstalls all of superpowers, not this one alone.
What separates a skill that changes behaviour from one that is ignored — mostly the description field.
Reach for it when Your team wrote skills and the model never triggers them.
npx skills add obra/superpowersInstalls all of superpowers, not this one alone.
One CLI, and it works across Claude Code, Codex, Cursor, Copilot, Windsurf, Zed and the rest — it writes into whichever agent directories it finds:
$ npx skills add addyosmani/agent-skills
The unit is the repository, not the skill. That command installs every skill in addyosmani/agent-skills — all of them, not just the one you read about. Worth knowing before you install dotnet/skills, which has 107. Install deliberately: context you do not need is context the model has to ignore.
Telemetry is on by default and feeds the registry's install rankings. DISABLE_TELEMETRY=1 turns it off.
addyosmani/agent-skills covers every phase above with one consistent vocabulary — it is the highest-coverage single install here. Add obra/superpowers for the execution discipline (TDD, plans, subagents, worktrees) and you have the spine of the lifecycle from two commands. Everything after that should be your stack: Expo for mobile, Trail of Bits for security work, the .NET set for .NET.
Nothing on this page is ours. Redline does not vendor, fork, host or modify any of these. Each entry links to the publisher's repository, and the install command pulls from them directly. If a publisher changes a skill, you get their change — this is a map, not a mirror.
Star counts appear on each entry as a dated snapshot: the number GitHub returned on the day we counted, printed with that date beside it. Install counts and audit verdicts are not here at all — we have no way to date those honestly, and a number that drifts with nothing to anchor it is worse than no number. 10 of the 15 publishers carry an Official badge, which means they ship the thing the skill is about — Vercel on React, Expo on Expo, the .NET team on MSBuild.
Skills shape what the agent writes. They do not enforce anything: no skill blocks a merge, and none of them is versioned against your repository or measured for whether it changed an outcome. That is the other half of the loop, and it is the half Redline does — see the output contract and the merge gate. Use both: skills to raise the floor on what gets written, the gate to catch what still gets through.