Rules for an AI coding agent to filter out generic AI-generated UI designs, text, and code.
JavaScript accessibilityagent-skillsai-agentsai-coding-agentai-slopanti-slopclaude-codecopywritingdesign-rulesui-design
v3.2.7: content round plus an OpenCode door (Two-State Layout, Over-Explained Comment, Decorative Status Dot, R-37 DESIGN.md conflict protocol, OpenCode verified)
miqdadbadzubair-design11dc5c14f994
diff --git a/.claude-plugin/contrast-mcp-launcher.mjs b/.claude-plugin/contrast-mcp-launcher.mjsindex de5cfc3..366601f 100755--- a/.claude-plugin/contrast-mcp-launcher.mjs+++ b/.claude-plugin/contrast-mcp-launcher.mjs@@ -1,12 +1,7 @@#!/usr/bin/env node/**- * Starts contrast-mcp.py with whichever Python name this machine actually has.- *- * The manifest's `command` is one string with no shell, so it cannot pick per- * platform on its own, and every single-name answer breaks somewhere: macOS has- * had no `python` since 12.3 removed Python 2, and Windows has no `python3` on- * some installs. v3.0.2 made allowed-tools and the skill prose accept both; this- * is the same fix for the one place that only takes a single value.+ * Starts contrast-mcp.py with whichever Python name this machine has.+ * The manifest's `command` is a single string and cannot pick per platform.*/import { spawn, spawnSync } from "node:child_process";import path from "node:path";@@ -17,9 +12,8 @@ const server = path.join(here, "..", "skills", "antislop-human", "contrast-mcp.pconst candidates = process.platform === "win32" ? ["python", "python3"] : ["python3", "python"];-// Run each candidate rather than just looking it up: Windows ships a `python`-// stub that exists on PATH and only opens the Store page, and pyenv leaves-// shims that exit non-zero when no version is selected.+// Run each candidate: the Windows `python` stub is on PATH but only opens the+// Store, and pyenv shims exit non-zero when no version is set.const python = candidates.find((bin) => spawnSync(bin, ["-c", ""], { stdio: "ignore" }).status === 0);diff --git a/.claude-plugin/marketplace.json b/.claude-plugin/marketplace.jsonindex 37d2c90..de11e13 100644--- a/.claude-plugin/marketplace.json+++ b/.claude-plugin/marketplace.json@@ -8,7 +8,7 @@{"name": "antislop","source": "./",- "version": "3.2.6",+ "version": "3.2.7","description": "Anti Slop: Rules for AI Coding Agents. Stops generic AI slop in generated UI, copy, and code. Loads as six skills."}]diff --git a/.claude-plugin/plugin.json b/.claude-plugin/plugin.jsonindex c187633..6bb0b28 100644--- a/.claude-plugin/plugin.json+++ b/.claude-plugin/plugin.json@@ -1,7 +1,7 @@{"name": "antislop","displayName": "antislop",- "version": "3.2.6",+ "version": "3.2.7","description": "Anti Slop: Rules for AI Coding Agents. Stops generic AI slop in generated UI, copy, and code. Loads as six skills.","author": {"name": "miqdadbadjuber"diff --git a/.codex-plugin/plugin.json b/.codex-plugin/plugin.jsonindex 62a0f61..517369e 100644--- a/.codex-plugin/plugin.json+++ b/.codex-plugin/plugin.json@@ -1,6 +1,6 @@{"name": "antislop",- "version": "3.2.6",+ "version": "3.2.7","description": "Anti Slop: Rules for AI Coding Agents. Stops generic AI slop in generated UI, copy, and code. Loads as six skills.","author": {"name": "miqdadbadjuber"@@ -13,7 +13,7 @@"interface": {"displayName": "antislop","shortDescription": "Anti Slop: Rules for AI Coding Agents.",- "longDescription": "Stops generic AI slop in generated UI, copy, and code. The core filter is always available; the UI, copywriting, human, mobile layout, and code comment skills load for the task at hand. A filter, not a style guide.",+ "longDescription": "Stops generic AI slop in generated UI, copy, and code. The core filter is always available; the UI, copywriting, human, responsive layout, and code comment skills load for the task at hand. A filter, not a style guide.","developerName": "miqdadbadjuber","category": "Productivity","brandColor": "#000000",diff --git a/.cursor-plugin/plugin.json b/.cursor-plugin/plugin.jsonindex 758b99f..9392226 100644--- a/.cursor-plugin/plugin.json+++ b/.cursor-plugin/plugin.json@@ -2,7 +2,7 @@"name": "antislop","displayName": "antislop","description": "Anti Slop: Rules for AI Coding Agents. Stops generic AI slop in generated UI, copy, and code. Loads as six skills.",- "version": "3.2.6",+ "version": "3.2.7","author": {"name": "miqdadbadjuber"},diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.ymlindex 5c58d17..6cd1143 100644--- a/.github/workflows/ci.yml+++ b/.github/workflows/ci.yml@@ -18,21 +18,16 @@ jobs:with:node-version: 20- # The rules antislop applies to other repos, applied to this one: the em- # dash ban with its own carve-out, rule and gate coverage, version drift,- # manifest validity, skill registration.+ # The rules antislop applies to other repos, applied to this one.- name: Repo guardrailsrun: node scripts/check-repo.mjs- # The contrast table in antislop-human is a set of claims about a formula.- # This recomputes every row. It caught issue #2 by hand once already.+ # The contrast table is a set of claims about a formula; recompute every row.- name: Contrast table matches the formularun: python3 skills/antislop-human/contrast-check.py --selftest- # skills/antislop/SKILL.md is generated from antislop.md. Editing the core- # and not syncing ships a skill that disagrees with the file it came from.- # sync-skills also throws on a raw.githubusercontent URL in the core, which- # is the Snyk W012 guard from v3.0.1.+ # skills/antislop/SKILL.md is generated from antislop.md, so an unsynced+ # core edit would ship a skill that disagrees with it.- name: Generated core skill is in sync with antislop.mdrun: |node cli/scripts/sync-skills.mjsdiff --git a/.gitignore b/.gitignoreindex 9f71a01..d0b2af1 100644--- a/.gitignore+++ b/.gitignore@@ -2,6 +2,9 @@docs/issue/+# Python bytecode, written whenever the contrast checker runs+__pycache__/+# OS noise.DS_StoreThumbs.dbdiff --git a/GUIDE.md b/GUIDE.mdindex 0e1db41..a66d4de 100644--- a/GUIDE.md+++ b/GUIDE.md@@ -44,7 +44,7 @@ npx antislop-aiIt shows you what antislop has, then asks three things:-- **Which agent(s) you use.** It detects the ones present in your project. Each agent reads antislop from its own folder, and the picker writes the pointer that reloads antislop in every session. Hermes installs globally (into `~/.hermes/skills/`) because it does not read project folders.+- **Which agent(s) you use.** It detects the ones present in your project. Each agent reads antislop from its own folder. On a project install the picker also writes the pointer that reloads antislop in every session; a global install writes no pointer and relies on the skills loading themselves by description. Hermes installs globally (into `~/.hermes/skills/`) because it does not read project folders.- **Which skills you want.** The core is always on. The extra skills load only for the work you do. See [What is a skill?](#what-is-a-skill).- **Where antislop should live.** This project only, or everywhere on your machine.@@ -127,6 +127,10 @@ The one-command way installs into the folder your agent reads. This is what it w| Gemini CLI | `.gemini/skills/` || Hermes | `~/.hermes/skills/` (global only) |+On a project install the picker also writes the pointer that reloads antislop every session: into the project's `AGENTS.md` for Codex, Antigravity, OpenCode, and Cursor, into `CLAUDE.md` for Claude Code, and into `GEMINI.md` for Gemini CLI. For OpenCode that is the whole mechanism: it loads the skill folders from `.opencode/skills/` and reads the pointer from `AGENTS.md`, which was verified against the opencode CLI.++Gemini CLI is legacy support: it was sunset on 18 June 2026 and superseded by Antigravity. The picker still installs into `.gemini/skills/` for existing Gemini CLI setups.+## What is a skill?A skill is an optional folder (with a `SKILL.md` inside) that goes deeper into one concern. The core works alone; a skill adds depth for one topic. Skills reference the core rules by number and never duplicate them, so adding one does not change the core.@@ -136,7 +140,7 @@ There are five skills. Pick the one that matches your work:- **UI work** (look and feel: layout, color, components, animation) → antislop-ui- **Copy work** (headlines, buttons, tone, made-up statistics) → antislop-copywriting- **People work** (readable colors, keyboard use, focus, button states) → antislop-human-- **Mobile layout work** (reflowing on a phone, tap targets, navigation) → antislop-layoutmobile+- **Responsive layout work** (reflowing across every screen width, phone to desktop, tap targets, navigation) → antislop-layoutmobile- **Code comments work** (remove generic AI comments, keep the valuable ones) → antislop-code- **More than one kind of work** → pick several. The picker lets you choose as many as you want.- **None of these** → fine. The core alone is a complete filter.@@ -147,6 +151,8 @@ The names above are what you pick in the one-command way and what the wizard askIt never beautifies on its own. antislop removes slop; it does not invent direction. If you have a specific look in mind, write it down in a file called `DESIGN.md` in your project and the AI builds toward it. You do not have to make one. Without a `DESIGN.md`, the AI labels its work "draft without direction" instead of passing it off as finished. A sterile result means the direction was missing, not that the filter failed.+If your `DESIGN.md` happens to ask for something antislop counts as slop, it does not obey quietly and it does not overrule you: it names the element, names the rule, and asks whether to keep it. Direction that is simply bold or unusual is not slop, and stays.+## Where is this going?antislop is packaged three ways at once: standard skill folders, native plugins for Claude Code, Antigravity, Codex, and Cursor, and the single-file core that works anywhere. Agent support grows over time. For the current release and what comes next, see the [roadmap](ROADMAP.md). For the full picture of every skill, see the [README](README.md).diff --git a/README.md b/README.mdindex c95a377..15cde48 100644--- a/README.md+++ b/README.md@@ -15,7 +15,7 @@> **Anti Slop: Rules for AI Coding Agents.** It stops them from generating generic "AI slop" UI and copy, without letting the result turn sterile. It is a **filter, not a style guide**: no prescribed colors, fonts, or layouts. It is not only for building pages: it also writes and audits copy, so AI text stops reading like AI. And it never beautifies on its own; `DESIGN.md` (yours) is where beauty and direction come from.-> **New here? Start with the [guide](GUIDE.md).** It explains what antislop is and how to install it, from zero.+> **New here? Start with the [GUIDE.md](GUIDE.md).** It explains what antislop is and how to install it, from zero.## What it does@@ -30,7 +30,7 @@ The core prevents slop but cannot invent direction. `DESIGN.md` (yours) suppliesantislop ships as a set of **standard agent skills** (one folder per skill, holding a `SKILL.md`). The core is always loaded; the other skills load only when the task needs them. Pick one of these six paths from this one repo.-**1. The picker (recommended).** One command, then answer the prompts. It asks where to install (this project or everywhere), which agents you use, and which extra skills you want, then copies the folders and writes the pointer that loads antislop every session. Path 2 does not write that pointer, so start here:+**1. The picker (recommended).** One command, then answer the prompts. It asks where to install (this project or everywhere), which agents you use, and which extra skills you want, then copies the folders. On a project install it also writes the pointer that loads antislop every session; a global install writes no pointer and relies on the skills loading themselves by description. Path 2 does not write that pointer, so start here:```bashnpx antislop-ai@@ -86,6 +86,8 @@ codex plugin add antislop@anti-slop| Gemini CLI | `.gemini/skills/` || Hermes | `~/.hermes/skills/` |+The Gemini CLI row is legacy support: Gemini CLI was sunset on 18 June 2026 and is superseded by Antigravity, which has full support. The picker still installs into `.gemini/skills/` for existing Gemini CLI setups.+The plugins (paths 3 to 6) are per-agent doors: they load antislop straight from this repo, so there are no skill folders to keep in sync.**Manual (single file, no packaging).** The core `antislop.md` alone is a complete filter you can paste into any chat window. Download it and tell your agent to read it; the First-Run wizard inside it installs skills the manual way:@@ -102,7 +104,7 @@ curl -o antislop.md https://raw.githubusercontent.com/miqdadbadjuber/anti-slop/m| antislop-ui | UI / visual: layout, color, components, decoration, motion, structure | v2.2.0 || antislop-copywriting | Copy & text: headlines, CTAs, tone, fake stats, anti-AI-writing patterns, markdown hygiene | v2.3.0 || antislop-human | Human: contrast (with the checker), keyboard, focus, states | v2.4.0 |-| antislop-layoutmobile | Mobile layout: responsive breakpoints, grids, overflow, tap targets, navigation | v2.5.0 |+| antislop-layoutmobile | Responsive / mobile: reflowing across screen widths (phone to desktop), breakpoints, grids, overflow, tap targets | v2.5.0 || antislop-code | Code comments: remove generic AI-slop comments, keep the valuable ones, never touch the code | v3.1.0 |Pick what matches the work:@@ -110,7 +112,7 @@ Pick what matches the work:- UI work → antislop-ui- Copy work → antislop-copywriting- People work → antislop-human-- Mobile layout work → antislop-layoutmobile+- Responsive layout work → antislop-layoutmobile- Code comments work → antislop-code- More than one kind of work → install several- None → the core alone is a complete filter@@ -136,6 +138,7 @@ What changed in each release. The full tracker, including the cross-agent plugin- **v3.2.4** turned R-35 into a click-through smoke test: every interactive element must be run and exercised one at a time, and its result recorded as evidence in the Delivery Gate report.- **v3.2.5** opened the Cursor door: `.cursor-plugin/plugin.json` plus a `.cursor-plugin/marketplace.json` index, skills with a `.mdc` rule pointer. The Codex plugin also gains its app identity: an icon, a brand color, and a banner screenshot.- **v3.2.6** merged a five-PR community round: the contrast MCP tool gets a Python launcher that works on macOS too, pointer writes stop touching the user's entry file, the smoke test can now fail (`npm test`), and the repo gains guardrail checks, a CI workflow, and a contributors section.+- **v3.2.7** is a content round plus an OpenCode door. Three slop patterns earned names: exactly two layout states (a phone stack and a desktop grid) with nothing between (`antislop-layoutmobile`), a comment that runs on for several lines around a one-line fact (`antislop-code`), and the decorative status dot beside a heading that glows and pulses while marking nothing (`antislop-ui`). Your `DESIGN.md` is no longer followed blindly: when it asks for a named slop pattern, the agent names the element and the rule and asks you to keep it or drop it (R-37). OpenCode is verified to load the skills from `.opencode/skills/` via its `AGENTS.md` pointer.## FAQ@@ -166,6 +169,14 @@ Thanks to everyone who helps make antislop better.Found a new AI slop pattern, a rule that missed something, or a bug in the installer? Open an [issue](https://github.com/miqdadbadjuber/anti-slop/issues). PRs are welcome for new AI slop patterns, clarifications, or checklist items out of sync with their rule.+<hr>++<p align="center"><em>“antislop is a filter, not magic.<br>+It clears the slop from your UI, text, and code.<br>+A beautiful UI is <code>DESIGN.md</code>'s job, and yours.”</em></p>++<hr>+## LicenseMIT: [LICENSE](LICENSE)diff --git a/ROADMAP.md b/ROADMAP.mdindex 2ddc8e1..6caa87e 100644--- a/ROADMAP.md+++ b/ROADMAP.md@@ -4,7 +4,7 @@## Where we are-The latest release is **v3.2.6**. antislop is a **packaged system**: a lean, always-loaded core plus five additive skills, each shipped as a standard agent skill folder (`skills/<name>/SKILL.md`). There are no per-skill READMEs.+The latest release is **v3.2.7**. antislop is a **packaged system**: a lean, always-loaded core plus five additive skills, each shipped as a standard agent skill folder (`skills/<name>/SKILL.md`). There are no per-skill READMEs.- **Core, unchanged since v2.2:** 38 rules across three tiers (Hard Gate, Purpose-Gate, Quality Locks), a Liveliness Toolkit, a mandatory Delivery Gate, and two usage modes (During / After). The single-file `antislop.md` is still a complete filter you can paste into any chat window; its First-Run Install Wizard is the manual install path.- **The six skills:** `antislop` (the core), plus `antislop-ui`, `antislop-copywriting`, `antislop-human`, `antislop-layoutmobile`, and `antislop-code`. Each skill only loads when the task needs it.@@ -43,20 +43,20 @@ That kept the filter pull-only-what-you-need and made the v3 packaging mechanica| v3.2.4 | R-35 sharpened into a click-through smoke test: every interactive element must be run and exercised one at a time, and its result recorded as evidence in the Delivery Gate report. || v3.2.5 | Cursor plugin door: `.cursor-plugin/plugin.json` manifest plus a `.cursor-plugin/marketplace.json` index, with the six skills and a `.mdc` rule pointer. The Codex plugin also gains its app identity (plugin icon, brand color, banner screenshot). || v3.2.6 | Five-PR community round (PRs #22 to #26): a runtime Python launcher for the contrast MCP tool (fixes it on macOS), pointer writes that leave the user's entry file untouched, a smoke test that can actually fail, repo guardrails plus a CI workflow, and a contributors section. |+| v3.2.7 | Content round plus an OpenCode door: three slop patterns earned names, "Two-State Layout" (a phone stack and a desktop grid with nothing defined between), "Over-Explained Comment" (a comment that runs on for several lines around a one-line fact), and "Decorative Status Dot" (a dot beside a heading that glows and pulses while marking nothing). R-37 gained the `DESIGN.md` conflict protocol: a direction that asks for a named slop pattern is surfaced and questioned, never silently followed or silently overridden. OpenCode is verified to load the skills from `.opencode/skills/` plus its `AGENTS.md` pointer. |## What's next-The plan is one item per version. The plugin doors follow the style of the Claude Code, Antigravity, Codex, and Cursor doors already shipped. v3.2.7 prepares the OpenAI listing.+The plan is one item per version. The plugin doors follow the style of the Claude Code, Antigravity, Codex, and Cursor doors already shipped.| Version | Item | What it means ||---------|------|---------------|-| v3.2.7 | OpenAI directory | Listing prep for the public plugin directory shared by ChatGPT and Codex: a versioned skill bundle, a submission document, and public privacy and terms pages. The submission itself happens on the OpenAI Platform after identity verification, outside this repo. |-| v3.2.8 | Gemini CLI | Gemini CLI was sunset in June 2026 and superseded by Antigravity CLI, which already has a door (v3.2.2). The door closes the folder anyway: manifest plus pointer. |-| v3.2.9 | OpenCode | OpenCode has no native plugin marketplace, so the door is the skills folder plus an `AGENTS.md` pointer (the picker already writes `.opencode/skills/`). |-| v3.2.10 | The rest | Hermes, GitHub Copilot, and the `.agents/skills` long tail (Cline, Roo, Amp). The existing `.agents/plugins/marketplace.json` already points at the repo root, so any agent that reads the .agents standard can add the same marketplace. Mostly verification plus documentation per agent. |+| v3.2.8 | The rest | Hermes, GitHub Copilot, and the `.agents/skills` long tail (Cline, Roo, Amp). The existing `.agents/plugins/marketplace.json` already points at the repo root, so any agent that reads the .agents standard can add the same marketplace. Mostly verification plus documentation per agent. |Beyond the numbered plan, with no promised version:+- **OpenAI public directory listing** (deferred). Listing prep for the plugin directory shared by ChatGPT and Codex: a versioned skill bundle, a submission document, and public privacy and terms pages. The submission happens on the OpenAI Platform after identity verification, outside this repo; parked until that can proceed.+- **Gemini CLI** (deferred). Gemini CLI was sunset in June 2026 and superseded by Antigravity CLI, which already has a door (v3.2.2). A completeness close would need an extension manifest for a tool that no longer ships; parked with antislop-compact.- **More plugin doors** as other agents grow plugin systems that antislop can ride from the same repo.- **antislop-compact**, a lightweight, standalone family of the five skills, each a self-contained cheat-sheet version that runs without the core alongside. Deferred: a parallel set of files would have to track every change to the full ones, and a compact skill is only sound if it keeps the one-line why per rule and the Delivery Gate.- **Skill candidates** still open: `antislop-docs` and `antislop-identity`.diff --git a/antislop.md b/antislop.mdindex a8fd6f9..4631ce4 100644--- a/antislop.md+++ b/antislop.md@@ -24,7 +24,7 @@ If no antislop pointer exists and this file is being read for the first time, ru- **2. `antislop-ui`** (UI / visual): pick this for building or editing a website, web app, or interface: color, layout, components, decoration, motion.- **3. `antislop-copywriting`** (copy & text): pick this for writing or editing copy: headlines, CTAs, value propositions, tone, landing-page text, product prose.- **4. `antislop-human`** (people): pick this for making sure a UI works for people with different eyes, hands, and setups: contrast, keyboard, focus, states.- - **5. `antislop-layoutmobile`** (mobile / responsive): pick this for layouts that have to hold up on a phone: breakpoints, scale, grids, overflow, tap targets.+ - **5. `antislop-layoutmobile`** (mobile / responsive): pick this for layouts that have to reflow across screen sizes, phone to desktop: breakpoints, scale, grids, overflow, tap targets.- **6. `antislop-code`** (code comments): pick this for writing or editing code comments: remove generic AI-slop comments, keep the valuable ones, never touch the code.- New skills appear here as they ship; never offer a skill that does not exist in this version.@@ -370,6 +370,9 @@ If an element genuinely cannot have a destination yet, remove it instead of ship- If no direction exists AND the user cannot be asked, the output MUST be labeled *"draft without direction"* AND use the honest default dials **ENERGY 1 / RHYTHM 1 / MOTION 1** (see Part 3). Never silently fall back to a neutral, sterile default- **FORBIDDEN**: designing without direction and silently falling into a neutral, sterile default- Style direction is the product owner's identity, not a slop pattern; this filter only applies on top of it+- If `DESIGN.md` itself asks for a named slop pattern, do not silently follow it and do not silently override it: name the element, name the rule it collides with, and ask the owner to keep it or drop it+- Ask only about a named pattern, never about a stylistic choice the direction is entitled to make: a bold palette or an unusual typeface is identity, not slop+- Record the answer in one line: if the owner keeps it, proceed and note the override; if they drop it, apply the rule- A design built without direction is a draft, not a shippable result#### R-38 — Real Content or Honest Placeholderdiff --git a/cli/index.mjs b/cli/index.mjsindex fe21a0b..c3f605c 100644--- a/cli/index.mjs+++ b/cli/index.mjs@@ -18,7 +18,7 @@ const EXTRA_SKILLS = [{ value: 'antislop-ui', label: 'antislop-ui', hint: 'UI and visual design rules' },{ value: 'antislop-copywriting', label: 'antislop-copywriting', hint: 'copywriting and text rules' },{ value: 'antislop-human', label: 'antislop-human', hint: 'accessibility, with the contrast checker' },- { value: 'antislop-layoutmobile', label: 'antislop-layoutmobile', hint: 'mobile layout rules' },+ { value: 'antislop-layoutmobile', label: 'antislop-layoutmobile', hint: 'responsive layout rules' },{ value: 'antislop-code', label: 'antislop-code', hint: 'code comment rules' },]@@ -34,7 +34,7 @@ function displayDir(agent) {async function main() {if (process.argv.includes('--version') || process.argv.includes('-v')) {- console.log('antislop 3.2.6')+ console.log('antislop 3.2.7')return}diff --git a/cli/lib/banner.mjs b/cli/lib/banner.mjsindex 2ee21dd..99cf5a6 100644--- a/cli/lib/banner.mjs+++ b/cli/lib/banner.mjs@@ -15,7 +15,7 @@ export function banner() {...LOGO,'',' ' + pc.dim('Anti Slop: Rules for AI Coding Agents'),- ' ' + pc.dim('installer v3.2.6'),+ ' ' + pc.dim('installer v3.2.7'),'',].join('\n')}diff --git a/cli/lib/install.mjs b/cli/lib/install.mjsindex 2eb2e61..1179971 100644--- a/cli/lib/install.mjs+++ b/cli/lib/install.mjs@@ -33,8 +33,7 @@ function resolveBase(location) {export function resolveTargets(location, selected = AGENTS.map((a) => a.id)) {const base = resolveBase(location)return AGENTS.filter((a) => selected.includes(a.id)).map((agent) => {- // A global-only agent (Hermes) always resolves to the home dir, even for a- // project install, because it never reads skills from a project folder.+ // A global-only agent always resolves to the home dir, even on a project install.const targetBase = agent.globalOnly ? os.homedir() : basereturn {agent,@@ -44,11 +43,8 @@ export function resolveTargets(location, selected = AGENTS.map((a) => a.id)) {})}-// Which agents already have their folder present, used to pre-check the picker's-// agent question. Absence of a folder is not absence of the agent, so the picker-// still lets the user add an agent whose folder does not exist yet. Global-only-// agents (Hermes) are detected only for global installs: a project install must-// not pre-check a folder that lives in the home dir.+// Agents whose folder already exists, used to pre-check the picker. A missing+// folder does not mean a missing agent, so the user can still add one.export function detectAgents(location) {const base = resolveBase(location)return AGENTS.filter((a) => {@@ -121,9 +117,8 @@ const SKILL_LINES = {'antislop-code': 'Code comments: `antislop-code`',}-// Names the skills instead of importing the core. The skills sit in the agent's-// own folder, so the agent already finds them; an `@` import would also pull all-// 46 KB of the core into every session, including sessions that touch no UI.+// Names the skills rather than importing the core: an `@` import pulls all 46 KB+// of it into every session, including ones that touch no UI.function pointerBlock(skills) {return [POINTER_START,@@ -144,9 +139,8 @@ function writeBlock(entry, block) {const head = replacing ? lines.slice(0, start) : lines.slice()const tail = replacing ? lines.slice(end + 1) : []- // Tidy the two seams only. Normalising the joined document, which is what a- // /\n{3,}/g sweep does, also collapses blank lines the author wrote, including- // the ones inside their fenced code blocks.+ // Tidy the two seams only: a /\n{3,}/g sweep over the whole document would also+ // collapse blank lines the author wrote inside their code fences.while (head.length && head[head.length - 1].trim() === '') head.pop()while (tail.length && tail[0].trim() === '') tail.shift()diff --git a/cli/package-lock.json b/cli/package-lock.jsonindex e0edfae..aab95ee 100644--- a/cli/package-lock.json+++ b/cli/package-lock.json@@ -1,12 +1,12 @@{"name": "antislop-ai",- "version": "3.2.6",+ "version": "3.2.7","lockfileVersion": 3,"requires": true,"packages": {"": {"name": "antislop-ai",- "version": "3.2.6",+ "version": "3.2.7","license": "MIT","dependencies": {"@clack/prompts": "^1.7.0",diff --git a/cli/package.json b/cli/package.jsonindex 020fd7e..dfd02da 100644--- a/cli/package.json+++ b/cli/package.json@@ -1,6 +1,6 @@{"name": "antislop-ai",- "version": "3.2.6",+ "version": "3.2.7","description": "Interactive installer for the antislop skills. Anti Slop: Rules for AI Coding Agents.","type": "module","bin": {diff --git a/cli/scripts/smoke-worker.mjs b/cli/scripts/smoke-worker.mjsindex 93b4b05..88bdaa4 100644--- a/cli/scripts/smoke-worker.mjs+++ b/cli/scripts/smoke-worker.mjs@@ -14,10 +14,8 @@ const skills = ['antislop', 'antislop-ui']const failures = []-/**- * Prints the line the script always printed, and remembers whether it held.- * Collecting instead of throwing means one run reports every break, not only- * the first, which is what you want from a smoke test you run after a refactor.+/** Prints the line this always printed and records whether it held, so one run+ * reports every break rather than stopping at the first.*/const check = (label, actual, expected) => {const ok = JSON.stringify(actual) === JSON.stringify(expected)@@ -25,13 +23,13 @@ const check = (label, actual, expected) => {if (!ok) failures.push(label)}-// 1. Fresh state: nothing detected, and the default selection is every agent.+// Fresh state: nothing detected, and the default selection is every agent.const fresh = detectAgents('project')check('A detected (fresh project)', fresh, [])const defaultTargets = resolveTargets('project')console.log('A default targets:', defaultTargets.map((t) => `${t.agent.id}@${t.path} exists=${t.exists}`).join(' | '), '(expect all seven, exists=false)')-// 2. Claude Code only, via explicit selection (old behavior preserved).+// Claude Code only, via explicit selection (old behavior preserved).const targets = resolveTargets('project', ['claude'])let written = installSkills({ skills, targets, overwrite: false })let pointers = updatePointers({ targets, skills })@@ -46,8 +44,7 @@ check('C written without overwrite', written.length, 0)written = installSkills({ skills, targets, overwrite: true })check('C overwritten', written.length, 2)-// 3. Antigravity on the fresh project: the reported bug. The target is included-// even though .agents/ did not exist, install creates it, pointer goes to AGENTS.md.+// Antigravity on a fresh project: .agents/ does not exist yet, install creates it.const agTargets = resolveTargets('project', ['antigravity'])console.log('D antigravity targets:', agTargets.map((t) => `${t.agent.id}@${t.path} exists=${t.exists}`).join(' | '), '(exists=false before install)')const agWritten = installSkills({ skills, targets: agTargets, overwrite: false })@@ -56,8 +53,7 @@ const agPointers = updatePointers({ targets: agTargets, skills })check('D antigravity pointers', agPointers.map((p) => path.basename(p)), ['AGENTS.md'])check('D .agents/skills/antislop/SKILL.md exists', fs.existsSync(path.join(process.cwd(), '.agents', 'skills', 'antislop', 'SKILL.md')), true)-// 4. New providers on the fresh project: OpenCode, Cursor, Gemini install and-// point to their own entry files.+// OpenCode, Cursor, and Gemini each install and point to their own entry file.for (const agent of ['opencode', 'cursor', 'gemini']) {const t = resolveTargets('project', [agent])const w = installSkills({ skills, targets: t, overwrite: false })@@ -67,23 +63,21 @@ for (const agent of ['opencode', 'cursor', 'gemini']) {check(`D2 ${agent} folder exists`, fs.existsSync(path.join(process.cwd(), agent === 'gemini' ? '.gemini' : agent === 'cursor' ? '.cursor' : '.opencode', 'skills', 'antislop', 'SKILL.md')), true)}-// 5. Hermes is global-only: even a project install resolves to the home dir, and-// it is never detected as a project agent. resolveTargets only, no install.+// Hermes is global-only: a project install still resolves to the home dir.const hermesProject = resolveTargets('project', ['hermes'])const hermesGlobal = resolveTargets('global', ['hermes'])check('D3 hermes project target', hermesProject[0].path, path.join(os.homedir(), '.hermes', 'skills'))check('D3 hermes global target', hermesGlobal[0].path, path.join(os.homedir(), '.hermes', 'skills'))check('D3 hermes detected in project', detectAgents('project').includes('hermes'), false)-// 6. Detection now sees the agents that were installed.+// Detection now sees the agents that were installed.const after = detectAgents('project')check('E detected after installs', [...after].sort(), ['antigravity', 'claude', 'cursor', 'gemini', 'opencode'])-// 7. Global with a selection.const globalTargets = resolveTargets('global', ['claude', 'codex'])console.log('F global targets:', globalTargets.map((t) => `${t.agent.id}@${t.path}`).join(' | '))-// 8. Copies are identical and the pointer block dedupes.+// Copies are identical and the pointer block dedupes.const src = fs.readFileSync(path.join(skillSourceDir(), 'antislop-ui', 'SKILL.md'), 'utf8')const dst = fs.readFileSync(path.join(process.cwd(), '.claude', 'skills', 'antislop-ui', 'SKILL.md'), 'utf8')check('G antislop-ui SKILL.md identical', src === dst, true)diff --git a/cli/scripts/sync-skills.mjs b/cli/scripts/sync-skills.mjsindex 773f661..70d82b6 100644--- a/cli/scripts/sync-skills.mjs+++ b/cli/scripts/sync-skills.mjs@@ -20,14 +20,17 @@ const coreBody = fs.readFileSync(path.join(repoRoot, 'antislop.md'), 'utf8').replace(/\r\n/g, '\n').trim()-// A shipped skill must not tell the agent to download the rest of its own instructions:-// that is Snyk W012, and it cost us one MEDIUM finding on the directory listing already.+// Snyk W012: a shipped skill must not tell the agent to download its own instructions.if (/https?:\/\/raw\.githubusercontent\.com/.test(coreBody)) {throw new Error('antislop.md carries a runtime download URL; remove it before syncing (Snyk W012)')}fs.writeFileSync(path.join(repoSkills, 'antislop', 'SKILL.md'), CORE_FRONTMATTER + coreBody + '\n')+// Python leaves a __pycache__ beside the contrast checker; it must not reach the tarball.fs.rmSync(cliSkills, { recursive: true, force: true })-fs.cpSync(repoSkills, cliSkills, { recursive: true })+fs.cpSync(repoSkills, cliSkills, {+ recursive: true,+ filter: (src) => path.basename(src) !== '__pycache__',+})console.log('Regenerated skills/antislop/SKILL.md from antislop.md and synced to cli/skills/')diff --git a/scripts/check-repo.mjs b/scripts/check-repo.mjsindex 090854a..f260916 100644--- a/scripts/check-repo.mjs+++ b/scripts/check-repo.mjs@@ -1,11 +1,7 @@#!/usr/bin/env node/*** The rules antislop applies to other people's repos, applied to this one.- *- * Every check here exists because the matching mistake already shipped: a gate- * item that disagreed with its rule (#7), checklists whose polarity no item used- * (#9), a contrast row that was wrong (#2), and the repo not passing its own- * filter (#6). Run it before opening a PR:+ * Run it before opening a PR:** node scripts/check-repo.mjs*/@@ -15,9 +11,8 @@ import { fileURLToPath } from 'node:url'const root = path.join(path.dirname(fileURLToPath(import.meta.url)), '..')const read = (p) => fs.readFileSync(path.join(root, p), 'utf8')-// Split on /\r?\n/, not '\n': on a Windows checkout the working-tree files carry-// \r\n, so a line kept its trailing \r and every ^...$ regex in the checks below-// (the em dash section tracker especially) silently stopped matching.+// Split on /\r?\n/, not '\n': on a Windows checkout a trailing \r survives and+// every ^...$ regex below silently stops matching.const lines = (p) => read(p).split(/\r?\n/)const SKILLS = fs@@ -36,11 +31,7 @@ const MANIFESTS = ['cli/package.json',]-/**- * R-02 bans the em dash, and R-02's own carve-out exempts the places that have- * to name the character to ban it. Encode that carve-out rather than skipping- * the check, so a new em dash in ordinary prose still fails.- */+/** R-02's carve-out, encoded so a new em dash in ordinary prose still fails. */function emDashes() {const bad = []const files = ['antislop.md', 'README.md', 'GUIDE.md', 'ROADMAP.md', 'SECURITY.md', ...SKILLS]@@ -99,8 +90,7 @@ function skillReferences() {/** The version is hand-written in eight places. They have to agree. */function versions() {- // A file that does not parse is already reported by the manifest check, so- // read it leniently here rather than crashing the whole run on it.+ // A malformed file is already reported by the manifest check; do not crash here.const json = (p) => {try {return JSON.parse(read(p))@@ -144,19 +134,14 @@ function frontmatter() {const bad = []for (const file of SKILLS) {const head = read(file).split('---')[1] ?? ''- // [ \t] not \s: \s spans the newline, so an empty value matches the first- // character of the next key and every skill passes.+ // [ \t] not \s: \s spans newlines, so an empty value would match the next key.if (!/^name:[ \t]*\S/m.test(head)) bad.push(`${file} has no name in its frontmatter`)if (!/^description:[ \t]*\S/m.test(head)) bad.push(`${file} has no description in its frontmatter`)}return bad}-/**- * A new skill folder has to be registered in four other places before anyone can- * install it: both rule pointers, the picker's menu, and the pointer block the- * picker writes. v3.1.0 had to touch all four to ship antislop-code.- */+/** A new skill has to be registered in all four places below or it cannot install. */function registration() {const names = SKILLS.map((f) => f.split('/')[1])const core = 'antislop'@@ -178,11 +163,7 @@ function registration() {return bad}-/**- * A rule's tier decides which Delivery Gate block it belongs in: Hard Gate to- * Block 1, Purpose-Gate to Block 2, Quality Locks to Block 4. A rule cited from- * the wrong block is what #7 was, three gate items disagreeing with their rule.- */+/** A rule's tier decides its gate block: Hard Gate 1, Purpose-Gate 2, Quality Locks 4. */function tiers() {const core = read('antislop.md')const part2 = core.split('## Part 2:')[1]?.split('## Part 3:')[0] ?? ''@@ -234,12 +215,7 @@ function manifestPaths() {return bad}-/**- * A rename like guide.md to GUIDE.md in v3.2.4 is exactly when a doc link goes- * stale, and existsSync cannot see it: macOS is case-insensitive, so the old- * link keeps passing locally and only breaks for readers on github.com. Compare- * against the real directory entry instead.- */+/** existsSync misses a case-only rename on macOS; compare the real dir entry. */const existsExactly = (target) => {const full = path.join(root, target)const dir = path.dirname(full)@@ -259,11 +235,7 @@ function docLinks() {return bad}-/**- * contrast-check.py and contrast-mcp.py each carry their own copy of the WCAG- * formula, and only the first has a selftest. The plugin exposes the second, so- * a drift between them would ship as a wrong answer with nothing to catch it.- */+/** The WCAG formula is duplicated, and only contrast-check.py has a selftest. */function contrastTwins() {const check = read('skills/antislop-human/contrast-check.py')const mcp = read('skills/antislop-human/contrast-mcp.py')diff --git a/skills/antislop-code/SKILL.md b/skills/antislop-code/SKILL.mdindex dddad43..3d5efdb 100644--- a/skills/antislop-code/SKILL.md+++ b/skills/antislop-code/SKILL.md@@ -68,6 +68,12 @@ allowed-tools: Read Write Edit Glob Grep## How It Should Read+### The Over-Explained Comment++- **Tell:** one comment that runs on for several lines, stacking reasons, context, and history around a fact that fits in one line: a four-line block explaining that a stub sits on PATH, which release introduced the workaround, and what broke before it. Every sentence is true. The length is the tell.+- **Why:** a person leaves a note, a generator writes a case. Padding a one-line fact into a paragraph, building a "because X, so Y, and therefore Z" chain, or citing the issue number and the version that fixed it are the same flourish as any other AI pattern, and they bury the one line that matters under the ones that do not.+- **Fix:** cut to the constraint alone: one line, two at most, never three. Keep the platform trap, the silent failure, the protocol rule, the performance cost. Drop the issue number, the version history, and the reasoning chain.+### Line-by-Line Narration- **Tell:** a comment on every trivial statement, narrating each line as it is written: `// Initialize count`, then `// Loop items`, then `// Get item`, then `// Increment`, then `// Return result`.@@ -104,6 +110,8 @@ Example that must stay:A comment earns its place when it explains something the code does not already show: the reason, the constraint, the non-obvious behavior.+Earning a place says what may stay, never how long it may run. A workaround note is one line about the workaround, not a paragraph about it. The example above is two lines because two facts are real, not because two lines is a target. This list is the most common reason a comment survives a review it should not: the content is legitimately valuable, so the length goes unexamined. Value is not length.+## Code Comment ChecklistRun these alongside the core Delivery Gate when the task touches comments. All answers must be **yes**:@@ -115,5 +123,6 @@ Run these alongside the core Delivery Gate when the task touches comments. All a- [ ] Do the comments avoid empty labels and vague TODOs that name no task?- [ ] Do the comments avoid decorative emoji and end markers?- [ ] Is the comment density one per logical block, not one per line?+- [ ] Is every comment one line, or two only when the second carries a new fact?- [ ] Do the remaining comments read short, natural, and in sentence case?- [ ] Is the scope guardrail held: only comments changed, the code untouched?diff --git a/skills/antislop-human/contrast-check.py b/skills/antislop-human/contrast-check.pyindex 11e553c..4c93693 100644--- a/skills/antislop-human/contrast-check.py+++ b/skills/antislop-human/contrast-check.py@@ -74,7 +74,10 @@ def reference_doc_path():def parse_reference_rows(path):- """Parse the reference table rows out of the skill doc."""+ """Reference table rows from SKILL.md, with the ratio already a float.++ A row that does not split into four cells is dropped silently, not raised.+ """rows = []with open(path, encoding="utf-8") as handle:in_table = Falsediff --git a/skills/antislop-human/contrast-mcp.py b/skills/antislop-human/contrast-mcp.pyindex bf2a8dc..26d5a36 100644--- a/skills/antislop-human/contrast-mcp.py+++ b/skills/antislop-human/contrast-mcp.py@@ -1,15 +1,14 @@#!/usr/bin/env python3-"""Minimal MCP stdio server that exposes the antislop contrast checker as a tool.+"""Minimal MCP stdio server exposing the antislop contrast checker as a tool.-Speaks the Model Context Protocol over stdio. All logging must go to stderr;-stdout is reserved for newline-delimited JSON-RPC responses.+All logging goes to stderr; stdout is reserved for JSON-RPC responses."""import jsonimport sysPROTOCOL_VERSION = "2024-11-05"SERVER_NAME = "antislop-contrast"-SERVER_VERSION = "3.2.6"+SERVER_VERSION = "3.2.7"TOOLS = [{diff --git a/skills/antislop-layoutmobile/SKILL.md b/skills/antislop-layoutmobile/SKILL.mdindex 7a143b1..6aeaaaf 100644--- a/skills/antislop-layoutmobile/SKILL.md+++ b/skills/antislop-layoutmobile/SKILL.md@@ -1,13 +1,13 @@---name: antislop-layoutmobile-description: "Mobile layout skill for antislop. Use for layouts that reflow on small screens: grids, overflow, tap targets. Load with the core."+description: "Mobile layout skill for antislop. Use for layouts that reflow across screen sizes, phone to desktop: grids, overflow, tap targets. Load with the core."allowed-tools: Read Write Edit Glob Grep---# antislop-layoutmobile> Anti Slop: Rules for AI Coding Agents. Mobile Layout skill-> Part of the antislop system. Read together with `antislop.md` (the core). This skill deep-dives the mobile layout concern: how a layout must reflow on small screens. Breakpoints, scale, grids, overflow, tap targets, and navigation. It references core rules by number and never duplicates or renumbers them. Load it when the task builds or edits a layout that has to hold up on a phone.+> Part of the antislop system. Read together with `antislop.md` (the core). This skill deep-dives the responsive layout concern: how a layout must reflow across screen sizes, phone to desktop. Breakpoints, scale, grids, overflow, tap targets, and navigation. It references core rules by number and never duplicates or renumbers them. Load it when the task builds or edits a layout that has to hold up at any screen width.## How to use this skill@@ -36,6 +36,12 @@ allowed-tools: Read Write Edit Glob Grep- **Why:** an override patch is not a mobile design. It fixes the symptom that got reported and leaves the next one, and the base styles stay tuned for a wide screen (R-03).- **Fix:** treat mobile as a designed state, not an override. Give the narrow viewport its own deliberate sizes and stacking, and verify the whole layout there, not just the patched spots (R-35).+### Two-State Layout++- **Tell:** a layout with exactly two states: a single stacked column below one breakpoint, and a wide multi-column grid above it, with nothing defined in between. A tablet or small laptop width then inherits whichever state is closest: the phone stack stretched absurdly wide, or the desktop grid crammed into a fraction of its intended canvas.+- **Why:** a page is a continuous range of widths, and R-03 demands it hold up at every one of them, not just two chosen breakpoints. Two states leave the whole middle band of the range (roughly 600 to 1024 px, where tablets and small laptops live) as an accident: content that neither stacks with intent nor sits in a grid that fits. The layout reads as designed only at its two sample points and broken everywhere between.+- **Fix:** define real states at the widths where the content stops working, and let them be as many as the content needs. A typical reflow is three states, not two: single column, then a two-column grid when cards get too wide as a single stack, then the full multi-column grid only when it genuinely fits. Verify by dragging the viewport through the whole range, not by checking two widths and calling it done (R-35).+## Scale & Sizing### Desktop-Sized Everything@@ -153,6 +159,7 @@ allowed-tools: Read Write Edit Glob GrepRun these alongside the core Delivery Gate when the task is mobile or responsive layout work. All answers must be **yes**:- [ ] Does the layout reflow into a distinct mobile state rather than a squeezed desktop? (R-03)+- [ ] Are there defined states across the width range, not just phone and desktop, so tablet and small-laptop widths are never a stretched stack or a crammed grid? (R-03, R-35)- [ ] Do sizes (type, padding, gaps, section heights) use a mobile scale, not desktop sizes unchanged? (R-03, R-05)- [ ] Do multi-column grids collapse and stack instead of colliding? (R-03)- [ ] Is there no horizontal overflow and nothing clipped? (R-03)diff --git a/skills/antislop-ui/SKILL.md b/skills/antislop-ui/SKILL.mdindex 109eb85..ef3c2c7 100644--- a/skills/antislop-ui/SKILL.md+++ b/skills/antislop-ui/SKILL.md@@ -179,6 +179,12 @@ allowed-tools: Read Write Edit Glob Grep- **Why:** the capsule-plus-glow-plus-dot combination is a self-referential badge that says "made by AI, about being made by AI". It adds noise, not information.- **Fix:** badges only when functionally needed, with the need written down, and never the full combination (R-09). A real status label is fine; a decorative "AI Powered" pill is not.+### Decorative Status Dot++- **Tell:** a small colored dot beside a heading, eyebrow, nav item, or label, usually glowing and pulsing on a loop, that marks nothing. It borrows the visual language of a live or recording indicator for a page where nothing is live.+- **Why:** the dot is an attention grab with nothing behind it: a glow plus an endless pulse is a double bid for the eye over a fact that does not exist. It reads as AI because generated pages reach for system-status vocabulary as decoration, and the same dot lands in the same place on every one of them.+- **Fix:** a dot must mark a real state (active, live, recording, warning). If it does, keep one dot, drop the glow, and drop the endless pulse (R-19). If it marks nothing, remove it: a heading needs no indicator to be a heading (R-31).+### Generic AI Typography- **Tell:** large monospace headings, or uppercase labels with extreme letter-spacing ("HOW IT WORKS", "FEATURES").@@ -290,6 +296,7 @@ Run these alongside the core Delivery Gate when the task is UI work. All answers- [ ] Does every navigation item and interactive element have a real destination or behavior, or a visible "Coming soon" label? (R-24, R-26)- [ ] Does motion follow the declared MOTION dial and serve a written purpose, with no endless loops? (R-19)- [ ] Is glass, glow, shadow, and radius used at their dose caps, not as a page-wide default? (R-10, R-11, R-12, R-13)+- [ ] Is every colored dot and status light marking a real state, with no decorative glow or endless pulse? (R-13, R-19, R-31)- [ ] On an app screen, is the layout built around the decision the user makes there, rather than the sidebar plus stat row plus chart plus table default? (C-3, R-20)- [ ] Is every number, delta, feed entry, and table row real or a labelled placeholder, with no invented metrics? (R-17, R-18, R-38)- [ ] Do empty form fields and table cells stay empty or carry honest placeholders (Your Name, [email protected]) instead of fake-looking data (John Doe, [email protected])? (R-23, R-38)diff --git a/skills/antislop/SKILL.md b/skills/antislop/SKILL.mdindex 859fdc8..e8fb4e5 100644--- a/skills/antislop/SKILL.md+++ b/skills/antislop/SKILL.md@@ -29,7 +29,7 @@ If no antislop pointer exists and this file is being read for the first time, ru- **2. `antislop-ui`** (UI / visual): pick this for building or editing a website, web app, or interface: color, layout, components, decoration, motion.- **3. `antislop-copywriting`** (copy & text): pick this for writing or editing copy: headlines, CTAs, value propositions, tone, landing-page text, product prose.- **4. `antislop-human`** (people): pick this for making sure a UI works for people with different eyes, hands, and setups: contrast, keyboard, focus, states.- - **5. `antislop-layoutmobile`** (mobile / responsive): pick this for layouts that have to hold up on a phone: breakpoints, scale, grids, overflow, tap targets.+ - **5. `antislop-layoutmobile`** (mobile / responsive): pick this for layouts that have to reflow across screen sizes, phone to desktop: breakpoints, scale, grids, overflow, tap targets.- **6. `antislop-code`** (code comments): pick this for writing or editing code comments: remove generic AI-slop comments, keep the valuable ones, never touch the code.- New skills appear here as they ship; never offer a skill that does not exist in this version.@@ -375,6 +375,9 @@ If an element genuinely cannot have a destination yet, remove it instead of ship- If no direction exists AND the user cannot be asked, the output MUST be labeled *"draft without direction"* AND use the honest default dials **ENERGY 1 / RHYTHM 1 / MOTION 1** (see Part 3). Never silently fall back to a neutral, sterile default- **FORBIDDEN**: designing without direction and silently falling into a neutral, sterile default- Style direction is the product owner's identity, not a slop pattern; this filter only applies on top of it+- If `DESIGN.md` itself asks for a named slop pattern, do not silently follow it and do not silently override it: name the element, name the rule it collides with, and ask the owner to keep it or drop it+- Ask only about a named pattern, never about a stylistic choice the direction is entitled to make: a bold palette or an unusual typeface is identity, not slop+- Record the answer in one line: if the owner keeps it, proceed and note the override; if they drop it, apply the rule- A design built without direction is a draft, not a shippable result#### R-38 — Real Content or Honest Placeholder