marcmarti9/agentit

using-agent-skills

Discover Agentit skills through bounded semantic packs, then let the primary AI choose only the concrete bodies that materially help the current stage or worker.

Langue du texte source : anglais

executive-board

Prepare board and investor decisions, narratives, pre-reads and governance communications around material metrics, variance, risks and explicit asks without spin or fabricated confidence.

Langue du texte source : anglais

executive-chief-of-staff

Triage company signals, maintain a decision/action queue, coordinate priorities and surface blockers with minimal noise while respecting authority and ownership.

Langue du texte source : anglais

executive-finance

Turn company financial data, unit economics, cash constraints and investment choices into explicit scenarios and executive decisions without pretending heuristics are universal rules.

Langue du texte source : anglais

executive-legal

Frame business legal and compliance decisions around issue, exposure, standard negotiating positions, jurisdiction and escalation without pretending AI advice is binding counsel.

Langue du texte source : anglais

executive-marketing

Make executive go-to-market, positioning, channel, funnel, brand and retention decisions by connecting customer evidence and channel economics to revenue.

Langue du texte source : anglais

executive-operations

Diagnose operational constraints, process/vendor risk, execution capacity and automation opportunities, then turn them into measurable operating decisions.

Langue du texte source : anglais

executive-orchestration

Orchestrate company-level decisions through a single accountable executive voice, selecting only the specialist executive skills and evidence that materially improve the current decision.

Langue du texte source : anglais

executive-people

Make executive people decisions about role design, hiring, compensation, performance, retention and organization structure while accounting for human and jurisdiction-specific risk.

Langue du texte source : anglais

executive-product

Make executive product decisions about customer problems, roadmap, product-market fit, sequencing, make/buy and investment using evidence rather than feature voting.

Langue du texte source : anglais

executive-strategy

Make company-level strategy decisions about positioning, markets, moats, priorities, partnerships and build/buy choices with explicit assumptions, evidence and trade-offs.

Langue du texte source : anglais

long-horizon-recovery

Keep substantial multi-session Agentit work resumable using private local state, compact checkpoints, and a fresh task decision when a new execution session resumes.

Langue du texte source : anglais

diagram-and-architecture-visuals

Choose and produce truthful diagrams for systems, processes, data, architecture and presentations. Route branded/general visual diagrams toward diagram-design, code-grounded interactive architecture maps toward Archify, and keep simple cases simple.

Langue du texte source : anglais

design-md-workflow

Read, create, maintain and verify a project's durable DESIGN.md visual-identity contract when persistent design tokens and rationale would materially improve multi-session frontend/design consistency.

Langue du texte source : anglais

anti-ai-slop-writing

Rewrite or review prose so it reads like a deliberate human writer rather than a generic chatbot. Preserve claims and citations, match the writer/brand voice, remove structural AI tells, and keep technical/factual text precise.

Langue du texte source : anglais

using-agentit

Lightweight provider-neutral entry point for material agent work. Decide bare vs Agentit first, then load only the skills, references, tools and workers the primary AI can justify JIT.

Langue du texte source : anglais

design-inspiration-research

Research current visual and interaction references before public greenfield or total-redesign art direction; extract design DNA, synthesize original directions, and keep provenance traceable.

Langue du texte source : anglais

mcp-tooling-fit

Select the smallest useful MCP/tool set for the current Agentit task, based on repository facts, explicit capability needs, least privilege, and current availability.

Langue du texte source : anglais

mobile-native-app-design

Build native-feeling Expo/React Native screens from real top-grossing app patterns. Use for mobile apps, onboarding, paywalls, tab bars, sheets, empty states, motion, or when the output must not look like a wrapped website. Prefer Appllama MCP when available;…

Langue du texte source : anglais

architect-orchestrator

JIT orchestration for material work. Help the primary AI choose useful topology, bounded workers, independent alternatives/review, and deterministic Loop/Graph execution without imposing a fixed hierarchy.

Langue du texte source : anglais

interview-me

Resolve only the material user-owned decisions that remain after inspecting available context. Batch independent questions, recommend defaults, challenge weaker methods, and preserve the user's final safe discretionary choice.

Langue du texte source : anglais

specialist-agent-routing

Select and use temporary domain specialists with bounded JIT context and concrete task-scoped skills.

Langue du texte source : anglais

marketing-and-growth

Growth engineering, conversion rate optimization (CRO), technical SEO, landing page copywriting, analytics, campaign/launch strategy, email, customer research, positioning, and content systems. Use for marketing, growth, launch strategy, SEO, or funnel…

Langue du texte source : anglais

reference-intelligence

Decide whether a task materially benefits from external references, then use the smallest domain-appropriate curated/live set. Distinguish authority, actually read material sources, distill useful knowledge into existing skills, and record durable provenance…

Langue du texte source : anglais

task-router

Compact AI-native decision protocol used after Agentit dispatch. The primary model chooses relevant semantic packs and any justified JIT skill/tool/reference set; cheap AI audits and strong AI arbitrates high-risk or unresolved disagreement.

Langue du texte source : anglais

planning-and-task-breakdown

Break clear requirements into ordered tasks. Use for multi-step work and dependencies; not for a single obvious edit.

Langue du texte source : anglais

debugging-and-error-recovery

Diagnose bugs and regressions from a red-capable feedback loop. Use for failing tests, builds, runtime failures, flaky behavior, or unexplained performance regressions.

Langue du texte source : anglais

verification-before-completion

Require fresh evidence before any done/fixed/passing claim. Under Agentit, also require the applicable Loop/Graph runtime receipt.

Langue du texte source : anglais

local-model-routing

Route work to local or remote models by capability tier. Use when local LLMs are available or the user wants local-first execution without silent quality drop.

Langue du texte source : anglais

ui-ux-pro-max-intelligence

Provider-neutral JIT UI/UX design intelligence adapter for the upstream UI UX Pro Max database. Use for product-specific style, palette, typography, UX, chart, icon, motion, landing, and stack guidance; do not treat it as the final creative director.

Langue du texte source : anglais

verification-gauntlet

Surround agent work with hard, signal-gated verification probes and anti-greenwash rules. Use before done claims, after implementations, or when tests look too easy; not for pure explanation.

Langue du texte source : anglais

design-trend-researcher

Research current design and creative-development trends, separating emerging patterns from mature conventions, saturated aesthetics, and speculative hype.

Langue du texte source : anglais

delight-and-whimsy

Add a restrained set of memorable, brand-appropriate interaction details, easter eggs, micro-moments, and playful affordances without sacrificing usability or turning the product into a gimmick.

Langue du texte source : anglais

visual-storytelling-director

Direct the pacing and narrative structure of premium web experiences using scene design, reveals, rhythm, continuity, interaction, and emotional beats before implementation.

Langue du texte source : anglais

creative-tool-scout

Research and select the best current creative-development tools, libraries, primitives, and implementation techniques for a chosen experience instead of defaulting to familiar stacks.

Langue du texte source : anglais

creative-web-experiences

Invent distinctive interactive web concepts before implementation. Use for premium redesigns, experiential sites, campaign pages, portfolios, product launches, editorial stories, or any request to make a site memorable without forcing a specific interaction…

Langue du texte source : anglais

threejs-spatial-experiences

Build premium navigable or guided 3D web environments such as stores, galleries, buildings, rooms, exhibitions, landscapes, and spatial brand journeys. Use when the experience is about moving through space rather than inspecting one product.

Langue du texte source : anglais

design-taste-frontend

Full-fat art direction for landing pages, portfolios, marketing sites, visual redesigns, and expressive frontend work. Reads the brief, selects a coherent visual world, fights AI defaults, and drives typography, composition, imagery, motion, and production…

Langue du texte source : anglais

emil-design-eng

Design-engineering craft inspired by Emil Kowalski's public skills: UI polish, component behavior, animation decisions, perceived performance, and invisible interaction details. Use for product UI, interaction polish, micro-interactions, motion critique, and…

Langue du texte source : anglais

figma-design-workflow

Use the official Figma MCP as design context and design-system authority for design-to-code, code-to-Figma, component mapping, variables, assets, and collaborative visual iteration.

Langue du texte source : anglais