Atlas / Skills / antfu / Anthony Fu's Skills

Anthony Fu's SkillsBLOCK

skills/antfu/antfu-design

Anthony Fu's curated collection of agent skills.

Verdict
BLOCK
Grade
D
Trust score
69 /100
Version
2026.9.25
Hosts
1 documented
License
MIT
Stars
5,992
01

Overview

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

A collection of Agent Skills reflecting Anthony Fu's preferences, experience, and best practices, along with usage documentation for the tools.

[!IMPORTANT] This is a proof-of-concept project for generating agent skills from source documentation and keeping them in sync. I haven't fully tested how well the skills perform in practice, so feedback and contributions are greatly welcome.

Installation

This repo mainly hosts the skills. The recommended way to use them is `@antfu/skills`, a skills pack installed with skills-npm. Besides my own skills, the pack includes curated skills from other authors, and your package manager keeps them in sync:

pnpm i -D skills-npm @antfu/skills
pnpx skills-npm setup

See antfu/skills-pack for the full list of skills in the pack.

Install From This Repo

To install skills from this repo directly, use the skills CLI:

pnpx skills add antfu/skills --skill='*'

or to install all of them globally:

pnpx skills add antfu/skills --skill='*' -g

Skills

This repo includes skills from different sources with different scopes.

Hand-maintained Skills

Opinionated

Manually maintained by Anthony Fu with his preferred tools, setup conventions, and best practices.

[!TIP] For design, see [antfu/design](htt
Read from source at commit 8d82d1932b46OBSERVED · 2026-10-21
Advertisement
02

Install

Commands as the repository documents them. They are shown, not run.

npm i -g vercel
npm i -g netlify-cli
npm i -D @pinia/testing
npm install
git clone --bare https://github.com/your-org/your-monorepo.git your-monorepo
npm install -D sass  # for .scss/.sass
03

Host compatibility

What the documentation claims. We have not run a compatibility test.

HostStatusNotes
cursormentioned
04

What it tells the agent

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: antfu-create-pr
description: Create a reviewable GitHub pull request from the current branch with a Conventional Commits title, a concise evidence-based body, and before/after screenshots for UI changes. Use when asked to open, create, publish, or prepare a PR.
metadata:
  author: Anthony Fu
  version: "2026.10.09"
---

# Create Pull Request

Open a PR that a reviewer can understand from the body alone. Explain the changed behavior and why, not the list of changed files. Keep the body proportional to the change: a one-line fix gets a few sentences, a cross-module feature gets tables and diagrams.

## Workflow

1. Inspect the repo: `git status`, current branch, remotes, default branch, `.github/PULL_REQUEST_TEMPLATE*`, and `AGENTS.md` / `CONTRIBUTING.md` for PR rules.
2. Compute the merge base against the target branch and record the exact `base...head` range the PR publishes. If the branch is stacked on another unmerged branch, target that branch and describe only this PR's own changes.
3. Read the full diff. Separate runtime code from generated files, lockfiles, and snapshots. Trace changed code to its entry points, callers, and external boundaries so architecture claims are backed by file paths or symbols.
4. Classify the PR (feature, fix, refactor, chore, docs) and decide which optional body sections earn their place. See [pr-body](references/pr-body.md).
5. Run the checks that match the changed surfaces (focused tests, typecheck, lint). Record the exact commands and results.
6. If the diff changes user-visible UI, capture before/after evidence. See [visual-evidence](references/visual-evidence.md).
7. Review the commit history and reshape it into atomic Conventional Commits. See [Commits](#commits).
8. Write the body to a temporary file. Push the branch. Create the PR with `gh pr create --title ... --body-file ...` (add `--attach` for each screenshot). Use `--draft` when checks are still running or the work is not review-ready.
9. Open the created PR and verify title, base, head, rendered tables, diagrams, and images.
10. Address review comments: fix each confirmed issue, run focused checks, push, reply with evidence, and resolve the thread.

## Title

Use Conventional Commits, matching how the repo already writes commit messages:

```text
feat(scope): add retry to upload client
fix(scope): avoid double submit on enter
refactor: extract diff parsing from cli
chore(deps): update vite to v8
docs: clarify worktree setup
```

- Lowercase, imperative, no trailing period, under ~70 characters.
- Scope is the package, module, or feature name the repo already uses. Omit it when the change is repo-wide.
- Squash-merge repos turn the title into the commit message; write it as the commit you want in history.

## Commits

Each commit does one logical thing and uses a Conventional Commits message. A reviewer should be able to read `git log --oneline` as the story of the change and review commit by commit.

```text
# Good: each commit is self-contained
feat(api): add task creation endpoint with validation
feat(ui): add task creation form component
feat(ui): connect form to api and add loading state
test(tasks): cover task creation (unit + integration)

# Bad: everything mixed together
feat: add task feature, fix sidebar, update deps, refactor utils
```

- Keep unrelated fixes, dependency bumps, and drive-by refactors out of the branch; open separate PRs for them.
- Before pushing, review the history and split or squash local commits (`git rebase -i <base>`) until each one stands alone and builds.
- Do not rewrite commits that are already pushed and under review; add new commits instead.

## Body

Required in every PR:

- `## Summary`: the problem or capability first, then what changed and why this approach. Mention if the PR is stacked and on what.
- Linked issues: `closes #123` / `fixes #123` / `refs #123` so GitHub links and auto-closes.
- `## Verification`: exact commands run and their results, plus anything left unverified.

Optional, only when it helps the reviewer:

- `## Change map`: table of module, before, after, reason. Use for changes across several modules or ownership boundaries. Never a raw file list.
- `## Architecture and behavior`: a Mermaid flow, sequence, or state diagram when ordering, async work, or state transitions changed. For a fix, pair a `Before` and `After` diagram at the same abstraction level.
- `## Boundaries and risks`: table of invariant, failure mode, protection, evidence. Use when there are real failure modes, migrations, or external effects.
- `## Visual changes`: before/after image table for UI changes.
- `## Rollout and follow-up`: migrations, feature flags, known gaps. State whether a gap blocks merge.

If the repo has a PR template, keep its headings and fill them; add sections above only where the template leaves room.

## Rules

- Say what is verified, what is assumed, and what is not verified. Do not write "safe", "fixed", or "backward compatible" without pointing at the evidence.
- Green CI proves only the checks CI runs. Focused tests prove only the behavior they cover.
- Do not describe inherited changes from a parent branch as this PR's changes.
- Use plain hyphens; never em dashes (U+2014) or en dashes (U+2013).
- No filler ("This PR aims to...", "comprehensive", "robust"). Start sentences with the fact.
- Do not paste tokens, secrets, user records, or full payloads into the body.
- Do not commit screenshots or other PR-only artifacts to the repo; attach them with `gh --attach`.
- If the PR was written with the help of an agent, say so in one line at the end of the body.
05

Trust audit

BLOCKgrade D · trust 69/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.

LayerWhat it checksResult
L0Provenance & inventoryPASS
L1Static analysis of the codeFAIL
L2Instruction surface (what it tells the agent)PASS
L3Class-specific surfacePASS
L4Behavioural (sandbox)SKIPPED

What the source does

Filesystem
none-observed
Network
declared (1 observation(s))
Shell
declared (4 observation(s))
Dependencies
not all pinned
Secrets in source
none-found

Findings (7)

HIGHCode injection · code.eval_exec · CWE-78, CWE-94, CWE-95
scripts/cli.ts:18
return exec(cmd, cwd)
Why it matters. evaluates text as code
Fix. remove; use a parser or a dispatch table
HIGHCode injection · code.eval_exec · CWE-78, CWE-94, CWE-95
scripts/cli.ts:53
exec(`git rm -f ${submodulePath}`)
Why it matters. evaluates text as code
Fix. remove; use a parser or a dispatch table
HIGHCode injection · code.eval_exec · CWE-78, CWE-94, CWE-95
scripts/cli.ts:147
exec(`git submodule add ${project.url} ${project.path}`)
Why it matters. evaluates text as code
Fix. remove; use a parser or a dispatch table
LOWInventory / provenance · inv.hidden_file · CWE-1104
.gitmodules
.gitmodules
Why it matters. hidden member outside the usual dotfiles
Fix. review its purpose
LOWSupply chain · supply.unpinned · CWE-829, CWE-1357
package.json
@antfu/eslint-config, @clack/prompts, @types/node, eslint, eslint-plugin-format, lint-staged, simple-git-hooks, skills-npm
Why it matters. 9 dependency range(s) float
Fix. pin exact versions or ship a lockfile
INFOPrompt injection · prompt.credential_read · CWE-94, CWE-1427
skills/nitro/references/core-configuration.md:143
- Use `useRuntimeConfig()` from `nitro/runtime-config` for env-overridable values; never read `process.env` in module top-level (edge runtimes only expose env during requests).
Why it matters. asks the agent to read credentials
INFOSupply chain · prompt.pipe_to_shell · CWE-829, CWE-1357
skills/pnpm/references/best-practices-ci.md:17
curl -fsSL https://get.pnpm.io/install.sh | sh -

Gates applied: no_behavioural_pass.

Audited 2026-10-21 · audit v0.4.1 · source sha 8d82d1932b46full audit observations/trust-audit/skill/antfu__antfu-design.json · Report an issue / request a re-scan
06

Audit history

Every audit this skill has had.

DateSourceVerdictGradeScoreChange
2026-10-218d82d1932b46BLOCKD69first audit
07

Questions

What does the antfu-design — antfu/skills skill do?

Anthony Fu's curated collection of agent skills.

Is antfu-design — antfu/skills safe to install?

No — not without reading the findings first. The audit graded it D (69/100) and found 3 critical or high issues in the source. Each one is listed on this page with the file and line it is on.

What can antfu-design — antfu/skills access on my machine?

The audit observed that it reaches the network and runs shell commands. Each of those is consistent with what it says it does. Secrets in the source: none found.

Which assistants does antfu-design — antfu/skills work with?

Its documentation mentions cursor. That is what the text claims, not a compatibility test we ran.

How current is this page?

The grade is for one exact copy of the source (8d82d1932b46), read on 2026-10-21. The repository is watched, and a new audit runs when it changes — this is the first audit.

Add this badge to your README

Shows this page's audit result and links back to it. It updates when the audit does.

ModelsAtlas audit badge

Markdown
[![ModelsAtlas audit: Anthony Fu's Skills](https://modelsatlas.com/skills/antfu/antfu-design/badge.svg)](https://modelsatlas.com/skills/antfu/antfu-design/)

Not your listing, or something out of date? Claim or update your listing.

Advertisement