NovusEdge

Design language for Claude artifacts and standalone HTML pages (reports, market research, explainers, dashboards, explorables, 3D). Use before building or editing any artifact or HTML page. Loads with one look: carbon, apple, primer, editorial, material, clay…

Architecture diagrams for polished-artifacts with LikeC4, built locally and published as an explorable page, an embedded view or an image. Use for system context, container and flow diagrams, and any architecture with more than one view or kept in a repo.

Apple HIG look for polished-artifacts, with no glass or translucency. Load after polished-artifacts:core when the user picks the Apple look. Spacious, grouped, content-first.

Editorial look for polished-artifacts, modelled on Tufte CSS and the AI 2040 Plan A scenario site. Load after polished-artifacts:core when the user picks the Editorial look. Ivory paper, et-book serif, one dark red accent, dot-fill figures.

IBM Carbon look for polished-artifacts. Load after polished-artifacts:core when the user picks the Carbon look. Flat, dense, productive; IBM Plex.

Clay look for polished-artifacts, modelled on anthropic.com's palette and type. Load after polished-artifacts:core when the user picks the Clay look. Warm ivory, serif body, bold sans headings, clay accent.

Manim (Community Edition) animations for polished-artifacts, rendered locally and embedded as video. Use only when the user asks for a Manim or 3Blue1Brown-style animation.

Material 3 look for polished-artifacts. Load after polished-artifacts:core when the user picks the Material look. Tonal surfaces, rounded containers, Roboto Flex.

Subagent orchestration for sophisticated opponent modeling and multi-party analysis.

Detect manipulation patterns, identify tactics being used against you, get counter-moves.

Quick reference for the 48 Laws of Power. Browse, search, or get specific laws.

High-functioning predator. Pattern recognition without sentiment. People are systems. Outcomes are all that matter.

Quick reference for the Art of Seduction. Seducer types, 24-step process, anti-seducer traits.

Self-terminating recursive orchestrator. Dispatches waves of subagents until a verifiable done-condition passes or the budget caps. Power with a kill-switch.

Quick reference for the 33 Strategies of War. Browse, search, or get specific strategies.

Multi-turn adversary simulation. Play out scenarios move-by-move with counter-moves.

Use before a change reaches its third file, when work will outlast this session or go to another agent, when resuming work an earlier session left uncommitted or unfinished, when the briefing or an edit hook says no feature is open, or before merging,…

Use when making a design choice, relying on a premise that may need review, leaving an open question, or before editing files that recorded decisions govern.

Use when writing or editing prose that feels padded, formulaic, promotional, or unlike its author, including technical docs, personal blogs, portfolio pages, comments, and PR descriptions. Use anti-slop-code for code structure.

Audit and trim code itself — narrator comments, defensive boilerplate, dead generality, tests that assert mocks, docstrings that restate the signature. Use for source files; use anti-slop for prose.

Use before writing a new function, class, component or module, and when a drygate message reports duplicated code. Covers finding existing code to reuse and removing a duplicate.

Sweep a codebase for functions that reimplement the standard library or an installed dependency, and for wrappers that add nothing. Haiku scouts flag candidates and Sonnet verifies each one. Run as /drygate:sweep [changed | paths...].

Review a repository's architecture from its design documents and decision ledger against the software-architecture reference, and write a dated review file under reviews/. Run as /software-architecture:review [paths]. Reviews design material, not code.

Use when designing or changing how a system is structured (service or module boundaries, data stores, events and messaging, durable workflows, caching, consistency, API or schema evolution, failure and overload handling, LLM integration), when writing an…

Strip machine-written tells from source code and technical documentation. Narrator comments, defensive boilerplate, dead generality, tests that assert mocks, docstrings that restate the signature.

Enforce clear technical prose — STE grammar, no AI buzzwords, no LinkedIn cadence. Default mode for code comments, docstrings, commit messages, PR bodies.

Use when the user asks for a new glowup pack, theme or look, wants a palette, color scheme, image, screenshot or mood turned into one, wants to change an existing pack, theme or color, or wants to change what glowup's status line shows, for example "make me a…

Use when making an animation, motion graphic, logo reveal, product shot, looping clip, launch video or GIF that is rendered from code (HTML/CSS, SVG, Canvas, WebGL/three.js, Remotion, GSAP, HyperFrames), when scoring one with synthesized sound, or when…

Use when writing, revising, or reviewing technical docs, specs, READMEs, procedures, or troubleshooting guides, or when asked to remove formulaic or AI-sounding prose from blogs and portfolio pages. Not a default style for unrelated conversation or creative…