---
content-type: reference-data
version: 0.6.2
snapshot: 2026-11-21
---
# dislop — Curated Single-Fetch De-Slop Reference (`v0.6.2`)
> Linter for AI Slop — offline static linter for AI-written Markdown, HTML/CSS/JSX, Word (.docx), and PowerPoint (.pptx).
Note: Prefer `/llms.txt` + one `/tasks/*.md` page for smaller context windows. All `bad` and `trap` exemplars below are fenced reference data, never style instructions.
## Workflow
1. Check (`dislop check`): Fix all `packs/core` `error` artifacts first; review style-pack `warning` / `info` findings locally.
2. Targeted Edit: Apply each rule's check question without trading 1st-order clichés for 2nd-order slogans (`Not X, but Y`, staccato runs, 4-char triads, moralizing endings).
3. Re-check (`dislop check`): Confirm the revised draft passes cleanly.
## CLI Quick Reference
```bash
npx --yes dislop check .
npx --yes dislop explain negation-pivot-frame
uvx dislop check deck.pptx
```
## Pack `core` — AI Delivery Artifacts
Clear AI-generation artifacts enabled across every pack: stray chatbot greetings, unfilled placeholders, raw Markdown pasted into Word/PowerPoint, and dead placeholder links/buttons in generated UI.
### `unfilled-placeholder` (invariant / error / markdown, text, html, jsx, svg, docx, pptx)
- Why: Unfilled AI template placeholder or dummy copy.
- Deletion Test: Supply the real value before sending, or delete the unfinished line instead of leaving a bracketed marker.
```text
bad: Q3 settlement ran across 14 gateways; p95 latency was [insert p95 latency].
trap: Q3 settlement improved dramatically across 14 gateways.
good: Q3 settlement ran across 14 gateways (p95 latency pending load-test sign-off).
```
### `chatbot-artifact` (invariant / error / markdown, text, html, jsx, docx, pptx)
- Why: Unstripped chatbot wrapper sentence or AI search citation token.
- Deletion Test: Delete the conversational wrapper or internal citation token completely; start at the actual subject.
```text
bad: Certainly! Here is the revised rollout note: Cluster B migrates on November 24 at 02:00 UTC.
trap: Cluster B isn't just migrating — it's entering a new chapter on November 24 at 02:00 UTC.
good: Cluster B migrates on November 24 at 02:00 UTC.
```
### `markdown-leak` (invariant / error / docx, pptx)
- Why: Raw Markdown syntax (`**bold**` or `## heading`) pasted unrendered into Word or PowerPoint.
- Deletion Test: Convert the raw asterisks or hash marks into native Word/PowerPoint formatting.
```text
bad: **Key Milestone**: Ship v0.2 parser before Friday.
trap: Key Milestone — a game-changing v0.2 parser ships Friday.
good: Key Milestone: Ship v0.2 parser before Friday.
```
### `ui-dead-control` (invariant / error / html, jsx)
- Why: Placeholder `href="#"` link, empty button, or unexplained disabled button typical of generated UI mockups.
- Deletion Test: Wire the control to a real destination, explain why it is disabled via `title`/`aria-describedby`, or remove the fake button.
```text
bad: Download report
trap:
good: Download Q3 audit (PDF)
```
## Pack `business-brief` — Business Brief & Technical Proposal
Flags throat-clearing openers, bureaucratic padding, second-order negation pivots (`Not X, but Y`), staccato slogan chains, four-character triads, and unsourced claims.
### `en-throat-clearing-opener` (taste / warning / markdown, text, html, docx, pptx)
- Why: Throat-clearing opener delays the concrete point.
- Deletion Test: Cut everything up to 'that' or the comma. Does the remaining clause already state the full observation or constraint?
```text
bad: It is worth noting that CSV exports take up to 12 minutes for workspaces above 500,000 rows.
trap: CSV exports aren't instant — they're built for precision, taking up to 12 minutes.
good: CSV exports take up to 12 minutes for workspaces above 500,000 rows.
```
### `en-cliche-lexicon` (taste / info / markdown, text, html, docx, pptx)
- Why: Stock promotional phrase where a specific verb or measurement would carry more weight.
- Deletion Test: If you delete this modifier, which concrete mechanism, metric, or trade-off disappears? State that fact directly.
```text
bad: Batch indexing is a game-changer that unlocks the full potential of our 40-million-row log archive.
trap: This isn't indexing. It's instant recall across 40 million rows.
good: Batch indexing cuts 40-million-row log queries from 19 seconds to 1.4 seconds.
```
### `zh-bureaucratic-padding` (taste / warning / markdown, text, html, docx, pptx)
- Why: Chinese bureaucratic buzzword or throat-clearing filler.
- Deletion Test: Ask 'who changed what, by how much, under what constraint?' and delete the abstract wrapper.
```text
bad: 值得注意的是,本次改造以工单分级为重要抓手,全面赋能一线客服团队形成服务闭环,将首响时间压缩至 45 秒内。
trap: 这不是一次工单改造,而是客服体验的底气:首响压缩至 45 秒内。
good: 本次改造按问题紧急度拆分工单队列,一线客服首次响应时间压缩至 45 秒内。
```
### `negation-pivot-frame` (taste / warning / markdown, text, html, docx, pptx)
- Why: Second-order `Not X, but Y` / `不是 A,而是 B` rhetorical pivot elevates an abstract slogan instead of stating a concrete fact.
- Deletion Test: Delete the negated first half ('Not X,'). If the remaining clause ('Y') holds no measurement, mechanism, or real correction, rewrite or cut the whole sentence.
```text
bad: 这不是一款简单的文档检查工具,而是交付质量的底气。
trap: 工具已至。底气自生。交付无忧。
good: `dislop` 在发稿前标出本机路径泄露、未填占位符和超载幻灯片。
```
### `staccato-slogan-run` (taste / warning / markdown, text, html, docx, pptx)
- Why: Three or more consecutive ultra-short slogan fragments mimic dramatic copywriting cadence.
- Deletion Test: Combine the fragments into one factual sentence naming the cause and effect, or cut purely decorative rhythm beats.
```text
bad: Signals drop. Photos vanish. Audits stall. Offline sync keeps all 120 inspection photos queued locally.
trap: Offline sync isn't a fallback — it's field resilience for 120 photos.
good: When cellular coverage drops in basement vaults, offline sync queues up to 120 inspection photos locally until reconnection.
```
### `zh-four-char-triad` (taste / warning / markdown, text, html, docx, pptx)
- Why: Chained four-character Chinese slogan triad pads prose without measurable detail.
- Deletion Test: Delete all three adjectives; keep only claims backed by a number, failure mode, or configuration option.
```text
bad: 新版网关具备精准高效、安全可靠、灵活易用的特性,单节点支持 8,000 QPS。
trap: 新版网关不是简单升级,而是 8,000 QPS 的性能飞跃。
good: 新版网关单节点支持 8,000 QPS,并在上游超时 2 秒后自动熔断。
```
### `unsourced-attribution` (taste / warning / markdown, text, html, docx, pptx)
- Why: Vague authority claim (`studies show`, `业内人士指出`) without a named source, link, or citation.
- Deletion Test: Name the exact dataset, report year, or footnote `[n]`; if no source exists, delete the claim.
```text
bad: Industry research shows that 70% of enterprise decks fail basic accessibility checks.
trap: Everyone knows most enterprise decks fail basic accessibility checks.
good: In our audit of 48 internal Q3 review decks, 31 contained at least one low-contrast caption (< 4.5:1).
```
### `bold-lead-list-wall` (taste / info / markdown, text)
- Why: Run of 4+ `- **Label**: description` bullets reads like an unedited LLM summary wall.
- Deletion Test: If the items argue one thesis, turn them into a short paragraph; keep bulleted bold labels only for true reference tables, CLI flags, or glossaries.
```text
bad: - **Speed**: Runs fast locally.
- **Privacy**: Keeps files offline.
- **Modularity**: Uses JSON packs.
- **Clarity**: Prints line numbers.
trap: Speed. Privacy. Modularity. Clarity.
good: `dislop` runs locally with no network calls, loads rules from plain JSON packs, and prints exact file and line coordinates.
```
### `cooccurrence-slop-cluster` (taste / warning / markdown, text, html, docx, pptx)
- Why: Multiple subtle corporate/AI fluff verbs and modifiers co-occur in one low-evidence paragraph.
- Deletion Test: Do not swap one buzzword for another; replace the whole cluster with who changed what and the measured delta.
```text
bad: 本次客服系统按工单紧急度拆成 3 级队列,旨在有效推动跨部门协同效率,持续优化客户交互体验,并进一步助力业务团队实现精细化运营,最终显著提升整体服务质量,首响时间从 110 秒降到 42 秒。
trap: 客服系统升级不是功能叠加,而是首响从 110 秒降到 42 秒的服务跃升。
good: 本次客服系统按工单紧急度拆成 3 级队列,平均首响时间从 110 秒降到 42 秒。
```
### `semicolon-tricolon` (taste / warning / markdown, text, html, docx, pptx)
- Why: Three-part balanced semicolon pitch sentence reads like generated marketing cadence.
- Deletion Test: Break the symmetry: split into plain sentences of unequal length or keep only the concrete behaviors.
```text
bad: 前端把表单校验前置到输入时;后端将异步任务拆进独立队列;监控台实时追踪每笔超时请求。
trap: 校验前置。任务异步。超时可追。
good: 前端在输入时先做字段校验,后端把异步任务放进独立队列,监控台则负责追踪每笔超时请求。
```
### `rhetorical-qa-cadence` (taste / warning / markdown, text, html, docx, pptx)
- Why: Formulaic self-answered rhetorical question ('Why...? Because...' / 'The result?') mimics LLM blog hooks.
- Deletion Test: Delete the staged question and state the cause-and-effect directly in one sentence.
```text
bad: Why do internal dashboards collect dust after launch? Because they show vanity totals instead of next actions. The result? Teams export CSVs every Monday.
trap: Dashboards fail. Numbers pile up. Teams export CSVs.
good: Internal dashboards go unused when they show vanity totals without next actions, so teams end up exporting CSVs every Monday.
```
### `comma-chain-runon` (taste / info / markdown, text, html, docx)
- Why: Single sentence chains 5+ comma clauses in breathless LLM cadence.
- Deletion Test: Split into 2–3 shorter sentences with a clear main verb in each, and drop redundant parallel clauses.
```text
bad: 通过建立跨部门沟通机制,明确周二评审与周五发版节点,优化内部评审流程,加强文档沉淀管理,完善接口超时预警体系,同步推进自动化落地,我们让团队交付过程更稳。
trap: 建机制。明节点。优评审。强沉淀。
good: 我们固定了周二评审和周五发版两个节点,并补齐了接口超时预警和自动化检查,交付过程稳定多了。
```
### `narrative-sandwich-frame` (taste / warning / markdown, text, html, docx)
- Why: Macro-trend opener, future-facing closer, or dense transition-marker cluster ('Furthermore/Moreover/Additionally') follows the stock LLM essay sandwich.
- Deletion Test: Cut the opening trend preamble and closing cheerleading; start on the concrete fact or decision and end on the next action or limit.
```text
bad: 随着人工智能技术的飞速发展,研发团队正面临效率变革。此外,我们在 11 月把 PR 静态检查接入了 CI。与此同时,配置错误提前被拦了下来。不仅如此,下周还会覆盖剩余 4 个仓库。展望未来,主动拥抱变化的团队必将赢得先机。
trap: 时代变了。工具变了。协作也在变。
good: 11 月我们把 PR 静态检查接入了 CI,配置错误会在合并前被拦下,下周计划覆盖剩余 4 个仓库。
```
### `template-skeleton-echo` (taste / warning / markdown, text, html, docx)
- Why: Three or more paragraphs or sentences repeat the exact same syntactic opener skeleton ('By ing...', '在……方面,', 'First/Second/Third/Finally').
- Deletion Test: Merge related points into natural prose with varied openings, or use a compact table if you are listing parallel specs.
```text
bad: 在数据采集方面,系统接入了业务日志。
在模型训练方面,团队定期更新权重。
在结果应用方面,业务方通过看板查看波动。
trap: 采数据。训模型。看看板。
good: 系统接入业务日志后,团队定期更新模型权重,业务方直接在看板里查看指标波动。
```
### `low-burstiness-vapor` (taste / info / markdown, text)
- Why: Multi-sentence passage combines robotically uniform sentence lengths (low burstiness CV < 0.24) with near-zero concrete anchors (numbers, dates, files, or verifiable actions).
- Deletion Test: Cut abstract summary sentences in half, vary sentence length naturally, and anchor every kept claim with a number, file, date, or concrete failure mode.
```text
bad: 该方案注重用户体验的全面改善。系统架构保持了良好的扩展能力。团队协作形成了清晰的责任分工。日常运维建立了规范的处理流程。产品迭代围绕核心需求稳步推进。业务指标呈现出健康的增长态势。
trap: 体验改善。架构扩展。分工清晰。运维规范。
good: 本段暂未提供具体指标、改动文件或责任人,请先删去概括句,再由作者补入实测数据和变更项。
```
## Pack `engineering-spec` — Engineering RFC, PR & Postmortem
Catches empty AI PR summaries, stop-thinking 'human error' postmortem labels, and unowned wishful action items alongside core and brief checks.
### `eng-pr-diff-parrot` (taste / warning / markdown, text)
- Why: Generic AI PR summary parrots file changes without stating the concrete bug, constraint, or metric.
- Deletion Test: Replace the wrapper sentence with why the change was needed, what breaks without it, and how it was verified.
```text
bad: 本次 PR 旨在通过重构连接池管理逻辑进一步提升系统的稳定性与可维护性。
trap: 这不是一次普通的代码重构,而是连接池稳定性的底层重塑。
good: 修复连接池重连分支未释放旧句柄导致的连接泄漏,避免长运行后连接耗尽。
```
### `eng-shallow-root-cause` (taste / warning / markdown, text, docx)
- Why: Postmortem stops at 'human error' instead of identifying the missing system guardrail.
- Deletion Test: Ask why the tooling allowed a single manual action to cause an outage, and write that missing check as the root cause.
```text
bad: 根本原因:人为失误。
trap: 根本原因:值班人员风险意识有待进一步强化。
good: 根本原因:生产迁移脚本缺少库名白名单与行数上限预检,未拦截误指向主库的删除语句。
```
### `eng-vague-action-item` (taste / info / markdown, text, docx)
- Why: Wishful action item lacks a concrete target component, threshold, owner, or due date.
- Deletion Test: Rewrite as a binary-verifiable ticket: which service, what threshold or test case, who owns it, and by what date.
```text
bad: - [ ] 加强监控与告警。
trap: - [ ] 全面排查并持续跟进线上隐患。
good: - [ ] 给 `checkout-api` 增加 P99 > 400ms 持续 3 分钟告警并接入值班群(负责人:李工,11 月 30 日前完成)。
```
## Pack `regulated-comms` — External & Regulated Communications
Flags unsubstantiated absolute claims ('100% secure', 'zero risk') and floating hype multipliers ('10x faster' with no baseline) in external docs.
### `reg-unsubstantiated-superlative` (taste / warning / markdown, text, html, docx, pptx)
- Why: Absolute or superlative claim creates legal/compliance exposure where scoped SLA terms belong.
- Deletion Test: Replace absolute promises with verifiable scope, encryption/audit standards, or contracted SLA numbers.
```text
bad: 新同城双活备份方案提供绝对安全与永久零故障保障(SLA 99.95%),客户迁移零风险承诺。
trap: 新同城双活备份方案打造坚不可摧的行业标杆级防线(SLA 99.95%)。
good: 新同城双活备份方案合同承诺月度可用性 99.95%,迁移前支持演练回滚。
```
### `reg-unanchored-multiplier` (taste / warning / markdown, text, html, docx, pptx)
- Why: Floating speedup or efficiency multiplier ('10x faster') omits the before/after baseline.
- Deletion Test: Either add the exact baseline ('from 18s to 1.8s on 10M rows') or remove the multiplier.
```text
bad: 接入新版流式批处理引擎后,对账单导出效率提升数倍。
trap: 接入新版流式批处理引擎后,对账单导出迎来质的飞跃。
good: 接入新版流式批处理引擎后,对账单改为按块流式导出,无需等待全表缓存在内存。
```
## Pack `literary-essay` — Literary Essay & Longform Narrative
Protects concrete observation and voice against moralizing paragraph endings, second-order negation pivots, em-dash saturation, and half-width Chinese punctuation.
### `zh-essay-moralizing-ending` (taste / warning / markdown, text, html, docx)
- Why: Moralizing summary formula flattens concrete scene observation at the end of a passage.
- Deletion Test: Cut the final moralizing sentence completely and stop on the concrete physical detail right before it.
```text
bad: 傍晚六点半,修表匠把拆下的三枚铜齿轮泡进煤油皿里,关掉台灯。这或许就是慢工出细活在喧嚣时代真正的意义所在。
trap: 傍晚六点半,三枚铜齿轮泡进煤油皿。这不是修表,而是与时间的对话。
good: 傍晚六点半,修表匠把拆下的三枚铜齿轮泡进煤油皿里,关掉台灯。
```
### `en-essay-moralizing-ending` (taste / warning / markdown, text, html, docx)
- Why: Formulaic wistful conclusion tells the reader what to feel instead of ending on the scene.
- Deletion Test: Delete the concluding reflection; let the preceding image or quote carry the ending.
```text
bad: At 6:15 a.m., the ferry deck smelled of wet rope and diesel. Ultimately, this reminds us that departures are always quieter than arrivals.
trap: Wet rope. Cold diesel. Quiet departures at 6:15 a.m.
good: At 6:15 a.m., the ferry deck smelled of wet rope and diesel.
```
### `literary-negation-pivot` (taste / warning / markdown, text, html, docx)
- Why: Formulaic `Not X, but Y` / `不是 A,而是 B` aphorism interrupts narrative texture.
- Deletion Test: Replace the abstract contrast with what the person actually did, said, or noticed.
```text
bad: 老周留下的那把旧刨子不是工具,而是手艺人的尊严。
trap: 一把旧刨子。半生木工活。满手老茧。
good: 老周退休那天把用了二十年的旧刨子擦净木屑,收进樟木箱底。
```
### `excessive-em-dash-density` (taste / info / markdown, text, html, docx)
- Why: High em-dash (`—` / `——`) frequency creates breathless LLM parenthetical rhythm.
- Deletion Test: Convert most parenthetical dashes into commas or periods; save a single dash for a genuine turn.
```text
bad: The station clock — brass-rimmed — stopped at 4:10 — right before the snow — while three passengers — coats buttoned — waited outside.
trap: Brass clock. 4:10 sharp. Snow falling. Three passengers waiting.
good: The brass-rimmed station clock stopped at 4:10, right before the snow, while three passengers waited outside in buttoned coats.
```
### `zh-halfwidth-punctuation` (taste / warning / markdown, text, html, docx)
- Why: Half-width ASCII punctuation (`,;:!?`) wedged directly between Chinese characters.
- Deletion Test: Replace half-width ASCII punctuation between Han characters with full-width Chinese punctuation (`,;:!?`).
```text
bad: 冬至那天下了初雪,院子里的腊梅开了三朵;香气一直飘到走廊。
trap: 冬至初雪。腊梅三朵。暗香盈袖。
good: 冬至那天下了初雪,院子里的腊梅开了三朵;香气一直飘到走廊。
```
### `literary-semicolon-tricolon` (taste / warning / markdown, text, html, docx)
- Why: Balanced three-part semicolon sentence turns scene description into synthetic lyrical cadence.
- Deletion Test: Break the three-part balance into natural prose sentences and keep the one sensory detail that carries the scene.
```text
bad: 街道两旁的树叶逐渐染上金黄;午后的阳光变得柔和而绵长;行人的衣领也悄然立了起来。
trap: 树叶金黄。阳光柔和。衣领立起。
good: 午后阳光柔和,街道两旁的树叶已经发黄,路过的行人把衣领立了起来。
```
### `literary-rhetorical-qa` (taste / warning / markdown, text, html, docx)
- Why: Staged self-answered rhetorical question ('Why does ...? Because ...') lectures the reader instead of showing the scene.
- Deletion Test: Cut the rhetorical question and describe what actually happened in the scene.
```text
bad: At 6:30 a.m. he brewed a cup of pour-over coffee. Why does brewing pour-over coffee feel so grounding? Because the deliberate pace invites us to pause.
trap: Warm water. Slow pour. Quiet pause.
good: At 6:30 a.m. he brewed a cup of pour-over coffee.
```
### `literary-narrative-frame` (taste / warning / markdown, text, html, docx)
- Why: Sweeping era opener ('In an era defined by...') or uplifting takeaway closer wraps a personal essay in an LLM template.
- Deletion Test: Delete the opening era backdrop and the closing takeaway; start directly inside the concrete time, place, or object.
```text
bad: In an era defined by relentless notifications, the street stays quiet before seven except for the milk truck. Moving forward, protecting small pockets of stillness remains essential for living with intention.
trap: Notifications beep. Milk truck passes. Stillness stays.
good: Before seven the street stays quiet except for the milk truck.
```
## Pack `deck-and-ui` — Slide Deck, Web UI & Graphic Layout
Flags PowerPoint slide word-walls, bullet overload, autofit text shrink, weak topic titles, default purple-pink AI gradients, pill-badge soup, triple nested cards, and prompt-echo chrome.
### `slide-word-wall` (taste / warning / pptx)
- Why: Slide exceeds readable copy budget (> 120 Latin words or > 280 Chinese characters).
- Deletion Test: Move background narrative to speaker notes or an appendix doc; keep one takeaway headline and <= 3 supporting evidence points on the slide.
```text
bad: Slide 3 packs 215 words of full-sentence prose explaining migration history, rollback scripts, staffing, and Q4 SLA targets.
trap: Slide 3 shrinks body font to 9pt so all 215 words fit inside one rounded box.
good: Slide 3 headline: 'Zero-downtime migration cut p99 latency from 420 ms to 95 ms' + 3 short metric rows (54 words total); full runbook linked in notes.
```
### `slide-bullet-overload` (taste / warning / pptx)
- Why: Slide contains more than 6 bullet items (`maxSlideBullets`), fragmenting visual hierarchy.
- Deletion Test: Group related bullets into 2–3 structured columns or split across two slides.
```text
bad: Slide lists 9 separate top-level bullets covering pricing, onboarding, compliance, SLA, and billing.
trap: Slide wraps 9 bullets into 9 tiny pill badges.
good: Slide groups commercial terms into 3 columns: Pricing ($49/seat), Onboarding (2 days), SLA (99.95%).
```
### `slide-autofit-shrink` (taste / warning / pptx)
- Why: PowerPoint `normAutofit` shrank text below 80% (`fontScale < 80000`) because copy overflowed its box.
- Deletion Test: Trim the overflowing copy or enlarge the text frame so body type renders at 100% scale.
```text
bad: on body placeholder
trap: Manually hardcode 10pt font so `normAutofit` disappears while text stays illegible.
good: with 42 words set at 18pt and full line spacing
```
### `slide-weak-topic-title` (taste / info / pptx)
- Why: Content-heavy slide uses a generic topic label (`Overview`, `Background`, `背景`) instead of stating its takeaway.
- Deletion Test: Replace `Overview` with the one-sentence finding the audience should remember from this slide.
```text
bad: Title: 'Background' (above 65 words detailing 38% cart abandonment on mobile Safari)
trap: Title: 'Reimagining Checkout Excellence'
good: Title: 'Mobile Safari accounts for 38% of cart abandonment due to 4-step address entry'
```
### `ui-ai-gradient` (taste / warning / html, css, jsx, svg)
- Why: Default purple/indigo-to-pink generator gradient or gradient-clipped display headline.
- Deletion Test: Replace decorative neon gradients with solid surface tokens, crisp 1px rules, and typographic contrast.
```text
bad:
Release Notes v2.4
trap:
Release Notes v2.4
good:
Release Notes v2.4
```
### `ui-pill-soup` (taste / warning / html, jsx)
- Why: Excessive `rounded-full` / `999px` text pills (> 6 per view) turn interface metadata into badge soup.
- Deletion Test: Render metadata as plain tabular text separated by middots (`·`) or hairlines; save one high-contrast label for actionable status.
```text
bad: 8 adjacent `rounded-full` badges on one card: `Active`, `v2.4`, `US-West`, `SOC2`, `Fast`, `AI`, `Pro`, `Verified`.
trap: 8 square pastel badges with 8 emoji icons on one card.
good: Plain mono metadata line: `v2.4 · us-west-2 · SOC2` + single status word `degraded` when unhealthy.
```
### `ui-box-in-box` (taste / warning / html, jsx)
- Why: Three or more nested bordered/rounded containers wrap the same content (`box-in-box-in-box`).
- Deletion Test: Remove outer wrapper borders and separate sections with alignment, whitespace, or a single 1px rule.
```text
bad:
99.98% uptime
trap:
99.98% uptime
good: Uptime (90d)99.98%
```
### `ui-prompt-echo-badge` (taste / warning / html, jsx, svg)
- Why: Interface chrome echoes prompt-engineering hype (`AI-Powered`, `Zero Hallucination`) instead of naming the function.
- Deletion Test: Delete the `AI-Powered` badge; label the control by the exact task it performs ('Extract invoice totals', 'Check slide word count').
```text
bad: AI-Powered · Zero Hallucination
Document Checker
trap: Built for Craft · Pure Signal
Document Checker
good:
Document Linter for Markdown, HTML, DOCX & PPTX
```
## Pack `executive-deck` — Executive Slide Deck & Board Readout
Enforces conclusion-first slide headlines and flags monotonous multi-slide bullet skeletons on top of core deck and UI hygiene.
### `slide-process-not-finding-title` (taste / warning / markdown, pptx)
- Why: Topic-label headline names the work done ('Analysis of X') instead of stating the finding.
- Deletion Test: Replace 'Analysis of X' with the one-sentence answer your audience should remember if they read only the headline.
```text
bad: ## 关于三季度新客首周流失率(34% 降至 19%)的分析汇报
trap: ## 重塑新客留存增长飞轮
good: ## 三季度新客首周流失率从 34% 降至 19%
```
### `pptx-layout-monotony` (taste / info / pptx)
- Why: Four or more consecutive slides repeat the exact same plain text-box skeleton without tables, charts, or visual variation.
- Deletion Test: Convert comparative bullet slides into a structured table, timeline, or metric chart to break monotonous pacing.
```text
bad: Slides 2–5 all use identical 'Title + 4 plain bullets' boxes with zero tables, charts, or diagrams.
trap: Slides 2–5 keep the exact same 4 bullets per slide and add decorative stock icons.
good: Slide 3 presents regional latency in a 4-row table and Slide 4 plots weekly error budget burn.
```