AGENT SKILL

design-audit

Static UI convention check - off-palette colors and wrong product-name casing, dishonest data labels (AI-generated or mock data shown as real), AI-slop (lorem remnants, hedge copy, emoji headings, uniform gradient grids), and client-facing language leaks. Reports file:line findings with fixes; conventions come from config layered over pack defaults. Use when the user asks to design-audit or check UI conventions, polish, or copy.

License 许可
MIT
最近更新
2026-09-28
Stars
★ 2
Forks
0

01 前言 Front Matter

name
design-audit
description
Static UI convention check - off-palette colors and wrong product-name casing, dishonest data labels (AI-generated or mock data shown as real), AI-slop (lorem remnants, hedge copy, emoji headings, uniform gradient grids), and client-facing language leaks. Reports file:line findings with fixes; conventions come from config layered over pack defaults. Use when the user asks to design-audit or check UI conventions, polish, or copy.
allowed-tools
Read, Grep, Glob, Bash(grep:*), Bash(ls:*)
argument-hint
[path | notes]

02 资源文件 2

03 SKILL.md 正文

/design-audit — conventions, honesty, and slop

A static pass over the UI's code and copy for the things reviewers skim past: colors off the palette, mock data wearing a real-data face, the tells of machine-generated filler, and internal language leaking to the user. Nothing is rendered — a rendered mode arrives with the browser probe (same deferral as /qa, decision 2026-07-27).

Adjacent skills: /design (the generative half — it produces UI, this inspects it; /design-audit stays purely detective and never generates) · /qa (behavior; /design-audit is look and language) · /audit code (correctness; /design-audit is convention) · /secure (exploitability; /design-audit flags leaked internals as language, not as a vulnerability).

<!-- acstack:runtime --> Run before the skill's steps — per invocation, not per session (4.36); failures degrade to markdown:

link="$(readlink "$HOME/.claude/skills/health" 2>/dev/null || true)"   # empty = not symlinked
pack="$(dirname "$(dirname "$link")")"   # NEVER trust this unless $link was non-empty
if [ "${link#/}" != "$link" ] && [ -x "$pack/bin/acstack-config" ] && ! "$pack/bin/acstack-config" runtime | grep -q '=off'; then
  "$pack/bin/acstack-config" || true          # resolved keys, with sources
  "$pack/bin/acstack-update-check" || true    # ≤1 fetch/day; silent ONLY if already checked today
  "$pack/bin/acstack-recall" || true          # LEARNINGS.md + bug-class names, capped 3KB
else
  echo "runtime off — proceeding without recall/update-check"
fi

<!-- /acstack:runtime -->

<!-- acstack:principles -->

Operating principles

  • Be direct. Push back in writing when the plan or the user is wrong. No sycophancy.
  • Never delete a decision. Supersede it: ~~old~~ → **Verdict (YYYY-MM-DD):** new call — reason.
  • Never fix, tune, or delete a test or eval case to raise a score. Log the miss honestly and leave the case unchanged.
  • Name exact things: regex patterns, function signatures, model names, before → after numbers. Never "fixed bugs".
  • Attribution: follow the project's attribution setting (default none) — no AI-tool mentions in generated docs, no attribution trailers in commits or PRs. Commit with explicit -m/-F messages only.
  • Config: read .claude/acstack.md at the project root (fall back to ~/.claude/acstack.md) before acting. ## Settings keys override pack defaults; a ## <skill-name> section overrides both. Unknown keys and sections are ignored.
  • Docs: BRIEF.md (frozen seed) / PLAN.md (living plan) / JOURNAL.md (rolling journal). If the repo uses legacy names (PLANNING_PROMPT.md / PLANNING.md / STATUS.md), use those instead — never create both.
  • Recall: if LEARNINGS.md exists at the project root, read it before starting.
  • Conduct: follow the acstack-conduct block in this repo's AGENTS.md — the word is the mode; the user sets the pace.
  • Hackathon lane: if the project's AGENTS.md carries the acstack:hackathon-lane block, only /do changes the repository during the event. Any other skill that would write a tracked file, commit or push says what it would have done and stops; a change that is not a task goes through the lane's operator route. <!-- /acstack:principles -->

Conventions come from config

Read a ## design-audit section from .claude/acstack.md:

  • palette: — the allowed hex values; colors outside it are findings.
  • banned-palette: — hex values that are findings wherever they appear, even if someone adds them to palette. Defaults to the violet-gradient family that generated UI reaches for (references/ai-tells.md).
  • product-names: — exact casings; other casings are findings.

Config always wins over the brand-neutral defaults in references/design-conventions.md. No config → use the defaults and say so in scope, naming what the reader is and is NOT getting: "no palette configured; flagged raw-hex sprawl and the dated default-look clusters (references/ai-tells.md Config), not brand conformance — nothing here knows what your brand is." Without a palette: there is no allowed set, so conformance is unanswerable rather than merely unchecked; saying only "flagged obvious sprawl" undersold both halves.

With palette: configured, the cluster check is suppressed: a look a project chose and declared is a decision, not a defect.

The four checks

Grep families and the default convention set live in references/design-conventions.md. Apply them with the Grep tool — this skill grants no shell git grep. No Grep tool? Use grep -rnE, never git grep.

  1. Palette + branding. Hardcoded colors outside the configured palette; product names in the wrong casing; spacing/font literals diverging from the project's own design tokens where tokens exist (a raw #3b82f6 beside a --color-primary token is the finding).
  2. Honest data labels. AI-generated, illustrative, sample, or mock data shown without saying so; a chart of fabricated numbers presented as real. The honest-measurement principle applied to pixels — the same stance /audit eval takes toward scores.
  3. Slop detection. Placeholder/lorem remnants, emoji-decorated headings in product UI, uniform gradient-card grids, hedge copy ("simply", "just", "seamlessly", "powerful", "effortlessly"), user-visible debug strings (console.log, TODO, FIXME in rendered text).
  4. Client-facing language. Internal jargon or codenames in UI strings; error messages exposing internals (stack traces, table names, file paths) — cross-referenced to /secure when they leak system detail; inconsistent terminology for the same object across screens ("order" here, "purchase" there).
  5. AI tells. The signature of generated UI — violet gradients, eyebrows, fabricated statistics, motion and materials violations, interaction-feel misses. Rules and greps in references/ai-tells.md.

Severity order is fixed: accessibility, then honesty, then everything else — regardless of how many hits each class produced. Unreadable text and a fabricated statistic are harm; a gradient is embarrassment. One tell is a choice; the full set is a signature, so say which it was.

No UI files in the path → say so and stop. If the target contains no markup, styles, or component files, there is nothing to audit; report the path, what was looked for, and stop. A "CLEAN" verdict over a directory with no UI in it is a true statement that reads as a false reassurance.

Stance and report shape

Read-only: findings and suggested fixes, never applied edits. First line is the verdict: CLEAN or <N> findings. Then:

  • Findings, grouped by the four checks: file:line · the convention violated · the suggested fix.
  • Safety checks: the exact greps run.
  • Scope: paths covered; static-only stated; whether a palette was configured or defaults were used.

04 同仓库更多 Skills

AaravChadha
AaravChadha/acstack

why

Decision archaeology - answers why the code, config, or plan is the way it is by searching the record in a fixed order: BRIEF constraints, dated PLAN decisions, JOURNAL entries, then git history. Stops at the first real answer, cites it with a date, and says no recorded rationale rather than inventing one. Use when someone asks why something is this way, why a choice was made, or where a decision is written down.

★ 2
AaravChadha
AaravChadha/acstack

verify

Audit a completion claim made by someone else - another session, agent, or teammate - against the running system rather than the diff. Re-derives what the claim's acceptance demands, runs it, and returns CONFIRMED, OVERSTATED, FALSE or UNVERIFIABLE with the command and its output. Never fixes, never edits the claim or the acceptance. Use when handed a "done", "all green", "merged" or "it works" you did not produce yourself.

★ 2
AaravChadha
AaravChadha/acstack

triage

Backlog hygiene sweep - stale items, duplicate pairs, missing acceptance criteria, unblocked-but-unassigned work, and milestone burn, delivered as a report first with only user-approved actions applied. Works on GitHub Issues in tickets mode and on PLAN.md checkboxes in document mode. Use when the user asks to triage, groom, or clean up the backlog or the plan's open tasks.

★ 2
AaravChadha
AaravChadha/acstack

ticket

Capture a brain-dump as a well-formed work item - verb-first title, acceptance criteria, file paths when known, an out-of-scope line. Files a GitHub issue in tickets mode or appends a numbered PLAN.md task in document mode; unknowns are marked TBD, never invented. Use when the user brain-dumps an idea, bug, or task to capture for later, or asks to ticket, file, or note something as work.

★ 2
AaravChadha
AaravChadha/acstack

ship

Branch-level release with five gates before the act - clean-state, tests, eval-vs-target, docs drift, and attribution sweep; any failing gate stops the release with its output. Then push, and under push: branch-pr open a report-shaped PR wiring the issue-closing reference in tickets mode or PLAN task IDs in document mode. Use when the user asks to ship, release, cut, or open the PR for a feature or branch.

★ 2
AaravChadha
AaravChadha/acstack

secure

Security review with a confidence gate - a finding exists only with a concrete exploit scenario and a high/medium/low confidence rating; everything else goes to a worth-hardening list, never inflated. Sweeps five surfaces: auth gates, secrets hygiene, injection and unsafe sinks, unsafe deserialization/crypto/transport, and LLM tool-use trust boundaries. Reports only, never fixes. Use when the user asks for a security review or to check vulnerabilities, secrets, or auth.

★ 2