ChloeVPin

AUDIT complet du dépôt pour la sur-ingénierie. Analyse l'ensemble du code au lieu du diff uniquement (comme ponytail-review). Liste classée de ce qu'il faut supprimer, simplifier ou remplacer par des équivalents de la bibliothèque standard/natifs. Rapport…

WHOLE-REPO audit for over-engineering. Scans entire codebase instead of diff (like ponytail-review). Ranked list of what to delete, simplify, or replace with stdlib/native equivalents. One-shot report, does NOT apply fixes. Trigger: "audit this codebase",…

AFFICHE l'impact mesuré de ponytail sous forme de tableau de bord compact : moins de code, moins de coût, plus de vitesse. Médianes de référence, pas de chiffres par dépôt. Affichage unique, pas un mode persistant. Déclencheurs : "/ponytail-gain", "ponytail…

DISPLAY ponytail's measured impact as compact scoreboard: less code, less cost, more speed. Benchmark medians, not per-repo numbers. One-shot display, not persistent mode. Trigger: "/ponytail-gain", "ponytail gain", "what does ponytail save", "show ponytail…

RECUEILLE chaque commentaire `ponytail:` dans le code source dans un registre de dette technique. Collecte les raccourcis délibérés et les reports pour qu'ils soient suivis au lieu de pourrir en "plus tard veut dire jamais". Rapport unique, ne modifie rien.…

HARVEST every `ponytail:` comment in the codebase into a debt ledger. Collects deliberate shortcuts and deferrals so they're tracked instead of rotting into "later means never". One-shot report, changes nothing. Trigger: "ponytail debt", "/ponytail-debt",…

VÉRIFIER le code d'interface pour la conformité aux Web Interface Guidelines. Récupérer les dernières directives depuis l'URL source, lire les fichiers spécifiés, appliquer toutes les règles, afficher les résultats au format concis `fichier:ligne`.…

REVIEW UI code for Web Interface Guidelines compliance. Fetch latest guidelines from source URL, read specified files, apply all rules, output findings in terse `file:line` format. Trigger: "review my UI", "check accessibility", "audit design", "review UX",…

Resolve conflicting claims, evidence, tests, or failures without forcing false consensus. Use when sources or observed behavior disagree and the coding agent must preserve uncertainty and decision boundaries.

Review dependency and supply-chain changes for provenance, vulnerabilities, integrity, execution behavior, and residual risk. Use when manifests, lockfiles, build scripts, or external artifacts change.

Compare a candidate coding change with a trusted baseline or reference to detect observable behavioral divergence. Use when ordinary tests pass but semantic alignment remains uncertain.

Keep coding-agent skills and guidance current by tracking provenance, freshness, deprecation, supersession, contradictions, and review triggers. Use when evidence, dependencies, versions, or failures change.

Verify that an AI-generated repository change matches the requested outcome, preserves unrelated behavior, and has adequate evidence before acceptance. Use for patches, commits, and pull requests.

Evaluate whether tests detect acceptance-relevant behavioral faults using boundaries, negative cases, differential checks, and mutation-oriented diagnostics. Use when tests are added, changed, weak, or contradicted by regressions.

Review AI-agent tools and proposed calls for least privilege, bounded parameters, explicit authorization, isolation, and auditable high-impact actions. Use when tools, credentials, scopes, or permissions change.

Review user-facing software changes for accessibility barriers across semantics, keyboard and other input, focus, structure, content, feedback, and assistive technology behavior. Use for web, mobile, desktop, CLI, and interactive UI changes; adapt the checks…

Activate for backend engineering work across APIs, services, data models, migrations, auth, validation, error handling, background jobs, queues, caching, rate limiting, observability, testing, performance, and safe production refactors. Use for…

Activate when a user requests a brand kit, brand identity board, logo concept, or visual identity system as an image output.

Activate when a user wants a high-contrast interface with hard edges, deliberate tension, and strict grid logic. Use for dashboards, technical portfolios, admin tools, command surfaces, or other interfaces that should feel sharp, direct, and usable.

Activate when visual quality is central to a frontend task and image generation is available — generate design reference images first, analyze them thoroughly, then implement the frontend from those references.

Activate for any task requiring a full, unabridged code or document output where truncation would render the result unusable.

Activate when a user needs a written design system reference that defines tokens, component rules, layout principles, motion, accessibility, and forbidden patterns for a project.

Activate when a user requests a polished, high-quality web interface — landing pages, SaaS dashboards, product pages, portfolios, or any frontend requiring strong visual craft and attention to detail.

Activate when a user requests a warm editorial-style interface, a document-like UI, or a strictly minimal aesthetic with near-zero color and heavy reliance on typography and whitespace.

Use when asked to audit, fix, or test web accessibility against WCAG 2.2 AA. Builds a matrix of states and viewports, runs axe-core, and requires evidence per row. Do not use for generic Lighthouse checks.

Use for OG images, social previews, and brand art. Generates text-free base artwork prompts and separate overlay specs. Prevents text leakage into image models. Do not use for a standalone mascot mark.

Use when building or reviewing gesture-driven UI, spring animations, and Apple-style interfaces. Enforces interruptibility, velocity-aware springs, and HIG principles.

Use when analyzing the X algorithm, explaining For You feed behavior, or optimizing posts. Maps retrieval, ranking weights, and visibility filtering to the open-source codebase. Read home-mixer/params/param.rs before answering.

Use when asked to generate a standalone mascot or symbol logo in a playful soft-3D style. Generates one mark without text, tiles, or app-icon containers. Do not use for full brand assets.

Use when asked to clean up, remove unused, or delete dead code. Finds unreferenced symbols, dedupes constants, and removes drift. Requires user approval before editing. Do not use for comments.

Use when writing new features or fixing bugs to enforce RED-GREEN-REFACTOR. Forces the agent to write a failing test, quote the failure, write implementation, and pass. Do not use for simple refactors.

Use when asked to clean up, remove dumb, or delete obvious comments. Flags comments that restate the code. Keeps why-comments and license headers. Do not use for dead code or commented-out code.

Quick reference for revanced-ponytail levels, skills, and stop phrases. One-shot display. Trigger: /revanced-ponytail-help, "ponytail help", "what ponytail commands", "how do I use revanced-ponytail".

Force the minimum code that meets real requirements. Use YAGNI, stdlib and native features first when they match the product, house style before personal style, rule of three for abstractions, root-cause bug fixes, and short skip notes. Levels: lite, full…

Whole-repo audit for over-engineering. Ranked list of what to delete, simplify, or replace with stdlib, native, or house patterns. Use when the user says "audit this codebase", "audit for over-engineering", "find bloat", "revanced-ponytail-audit", or…

Harvest every revanced-ponytail or ponytail debt marker into a ledger so deliberate shortcuts do not rot. Use when the user says "ponytail debt", "revanced-ponytail-debt", "/revanced-ponytail-debt", "what did we defer", or "list the shortcuts". One-shot…

Show measured impact from the upstream agentic benchmark only: less code, less cost, more speed, with limits. One-shot display. Not a per-repo savings number. Trigger: /revanced-ponytail-gain, "ponytail gain", "what does ponytail save", "show impact…

Review a diff for over-engineering only. Find what to delete or replace with stdlib, native, or house patterns. One line per finding. Use when the user says "review for over-engineering", "what can we delete", "simplify review", "revanced-ponytail-review", or…

Full-output enforcement for AI coding agents. Blocks truncated files, placeholder stubs, and soft exits. Use when the user needs complete implementations, full files, or exhaustive multi-part deliverables.

Strict frontend rails for AI coding agents. Concept-first craft for landings and product marketing sites. Forces a one-line concept, owned visual system, conversion jobs in order, signature moment when ambition is high, anti-slop defenses, and a review…

Audit-first redesign skill for existing sites and apps. Scans the codebase, lists design failures, then upgrades type, color, layout, motion, and states without breaking routes, forms, or brand locks. Use when improving an existing project rather than…

Generates a project DESIGN.md (semantic design contract) for handoff to design tools and coding agents. Captures atmosphere, color roles, type, components, layout, motion, responsive rules, and anti-patterns. Use when the user wants a DESIGN.md for Google…

High-pressure frontend rails for models that under-vary layout and over-wrap headlines. Forces a written design plan, AIDA page flow, hard hero line limits, gapless feature grids, stronger motion discipline, and aggressive anti-slop bans. Use when default…

Design and implement a distinctive, award-caliber web or product interface from a written brief without a reference image. Use when Codex must create a landing page, portfolio, SaaS site, product experience, developer tool, care or education interface,…

Design and critique user interfaces through task clarity, hierarchy, interaction states, accessibility, and implementation-aware decisions.

Reconstruct a web or product interface from one or more attached reference images with high visual fidelity. Use when Codex must copy, reproduce, rebuild, or closely match a screenshot, mockup, landing page, portfolio, dashboard, mobile screen, or visual…

Guidance for writing and refining prompts, tool descriptions, and agent instructions for GPT-5.6 workflows.

Shape agent replies for ADHD or high executive-function load: action-first first line, four-item chunks, visible Parked list, anchored time estimates, default-over-menu, depersonalized errors, less scaffolding in hyperfocus. Use when the user wants…

Apple's approach to interface design and fluid, physical motion, translated for the web. Revamped from Emil Kowalski's apple-design skill (MIT). Use when building or reviewing gesture-driven UI, spring animations, drag/swipe/sheet interactions, momentum and…

Flag comments that only restate the code and remove only the ones the user approves. Use when asked to remove dumb, redundant, or obvious comments, clean up comments, or invoke /remove-redundant-comments [<number>|all].

Audit a codebase for dead or otherwise safe-to-remove code and report every finding for approval before editing. Use when asked to find dead or unused code, deduplicate repeated constants, replace magic literals that duplicate a named constant, or remove…

Reformulate an existing video script into a clear, didactic, deep script for an entrepreneur audience. Use when asked to rewrite, reformulate, or modernize a video script, including Portuguese-language content for business owners.