Experience OS Agents Skills Directory

Experience OS agents perform their work through specialized skills. Each skill is designed for a specific task and is assigned to the agent with the relevant expertise. Skills are context-aware and appear in relevant areas throughout Experience OS based on the task you're performing.

This directory provides a complete list of available skills, including what they do, where they are available, and which agent uses them.

General skills

Ask anything (knowledge base)

• What it does: Led by the personalization expert agent, this skill provides instant answers to product and configuration questions directly from our official documentation. Instead of searching through articles yourself, the agent surfaces a concise, accurate answer. It draws from two official sources: Support articles from this knowledge base and our Developer Docs.
• Where it's available: Accessible from the central chat panel and the help icon (?) in the top navigation of Experience OS.
• Why it matters: Eliminates the time spent digging through documentation tabs, enabling you to quickly troubleshoot setup, configuration, reporting, or feature mechanics without leaving your current workspace.
• Specifications: 
  In-scope: Questions regarding setup, configuration, troubleshooting, reporting, feature capabilities, and statistical calculations (such as Probability to Be Best).
• Tips for best results:
  • Be specific about the feature or application you are working in (for example, mention whether you're in Experience Web, Reconnect, or the Experience App) so the agent can bypass clarifying questions.
  • Describe what you're trying to achieve rather than just entering a lone error code or keyword (for example, "How do I set up a feed for email recommendations" yields a better result than just "feed setup").

Creation skills

Generate template variables

• What it does: Led by the developer agent, this skill automatically scans your campaign’s HTML, CSS, and JS code to identify hardcoded values. Upon your approval, it rewrites the code using semantic variable syntax and syncs everything to the Variables tab in under 60 seconds.
• Where it's available: In the variation editor on the template screen in the following apps: Experience Web, Experience Email, and Personalization Breeze.
• Why it matters: Turning custom code into reusable marketer templates typically requires developer resources. This skill removes that bottleneck, enabling marketing teams to quickly change text, colors, images, or configurations without touching code.
• Specifications:
  • Supported variable types: Text (visible content), image (source URLs), color picker (HEX/RGB codes), number values
  • Organization: Automatically groups variables into logical layout structures (for example, "Hero Section", "Product Strip", "Footer CTA") for easy navigation.
  • Control: You can review, edit, or explicitly remove any suggested variables before clicking Implement All.

Generate new design

• What it does: Led by the designer agent, this skill builds production-ready layout and copy variations based on descriptions or strategic goals, while the developer agent handles the underlying clean, responsive code.
• Where it's available: In Experience Web Dynamic Content campaigns.
• Why it matters: Reduces execution time by transforming a text-based campaign brief into a production-ready campaign variation that you can review and edit directly in Experience OS.
• Specifications:
  • In-scope: Generates full visual layouts, components, and matching copy variations.
  • Control: The design is presented visually inside the chat panel, where you can request adjustments and apply it directly to the editor for further manual modifications.

Translate variation content

• What it does: Led by the copywriter agent, this skill translates all text variables within a campaign from your default language into one or multiple target languages in a single batch session. Context-aware AI references your Brand Book and campaign goals rather than executing literal word-for-word translations.
• Where it's available: Variation and template editing pages for any campaign type that has multi-language support enabled.
• Why it matters: Manually copying dozens of variables back and forth into external translation tools is slow and error-prone. This replaces a 30-minute chore with an instant localized workflow that respects your global brand guidelines.
• Specifications:
  • In-scope: Batch translation of all visible text variables into any locale currently configured in your section settings.
  • Regional support: Factors in localized variants (for example, distinguishing between Spanish for Spain (es_ES) vs. Spanish for Mexico (es_MX)).
  • Out-of-scope: Can't translate to languages that aren't defined in your Dynamic Yield section settings (the agent will direct you to configure them there if necessary).

Validate variation accessibility (accessibility reviewer)

• What it does: Led by the developer agent, this skill reviews your campaign code for WCAG 2.1 AA compliance and provides clear before/after code comparisons with one-click implementation options.
• Where it's available: In the Campaign Editor for any campaign that uses HTML, CSS, or JS. It can be triggered manually from the skills menu or directly from a chat conversation.
• Why it matters: Web accessibility compliance is legally mandatory for many brands and important for corporate reputation. This skill integrates expert audit capabilities natively into your creation flow, removing the need for separate developer compliance tools.
• Specifications:
  • Audit scope: Checks for missing or inadequate image alt text, insufficient color contrast ratios, missing keyboard navigation support, and ARIA attribute violations.
  • Output: Returns a prioritized list of up to 10 identified issues sorted by severity, complete with plain-language explanations of why each item matters.

Analysis skills

Analyze A/B test results

• What it does: Led by the analyst agent, this skill deep-dives into any active or completed A/B test report to provide a polished, concise executive summary highlighting primary performance metrics, statistical thresholds, and actionable conclusions.
• Where it's available: You can access it directly from any A/B test report page.
• Why it matters: Interpreting raw mathematical charts and documenting insights for stakeholders can require a great deal of manual effort. This skill automatically translates performance metrics into a presentation-ready narrative.
• Specifications:
  • Calculations considered: Highlights primary metric uplifts, statistical significance, Probability to Be Best percentages, secondary metrics support data, and projected real-world business gains (for example, projected monthly purchases).
  • Sharing: Includes a one-click option to download the full summary report for quick distribution or inclusion in decks.

Generate A/B executive report (experimentation program overview)

• What it does: Led by the analyst agent, this skill runs an aggregate analysis across your account's entire testing history to evaluate the comprehensive health, velocity, and performance of your optimization team.
• Where it's available: Available from the main Experience OS dashboard or can be invoked from the chat at any time.
• Why it matters: Essential for leadership reviews, quarterly business reviews (QBRs), and high-level program health tracking without needing manual data exports or spreadsheet models.
• Specifications:
  • Monitored metrics: Testing velocity (launches month-over-month), active variations count, learning rate (percentage of tests reaching conclusive results), time-to-conclusiveness efficiency, win rate against control, launch cadence by day of the week, and test-by-test takeaways.
  • Reporting: Evaluates parameters over a trailing 12-month window against baseline industry benchmarks to highlight structural opportunities.

Optimize Personalization Impact report setup

• What it does: Led by the analyst agent, this skill audits your Personalization Impact report configuration against every live campaign in your section. It surfaces misaligned experiences (such as losing variations, non-personalized baseline content, or inconclusive experiments) and recommends adjustments.
• Where it's available: On the Personalization Impact report page (or through a proactive alert if the system detects configuration anomalies). The skill can also be invoked from the chat at any point.
  • This skill is only available in sections with a Personalization Impact report. To activate this report in your account, contact your CSM.
• Why it matters: The Impact Report is your primary tool for proving the ROI of personalization to leadership. Incorrectly including broken or underperforming tests can artificially drag down your overall impact figures, leading to time-consuming manual audits.
• Specifications:
  • Evaluation parameters: Reviews every campaign by duration, type, targeting logic, primary goal, Probability to Be Best, and various other data points.
  • Interactive output: Provides a program status classification (for example, Early Stages, Building Phase, Strong Foundation), maps functional gaps (like missing funnel coverage or lack of campaign diversity), and isolates exact experiences to add or exclude with concrete explanations.

Investigate alerts

• What it does: Operating as a nightly automation service, this skill is led by the analyst agent. It scans your active production environments for technical errors and section-level performance drops, wraps these issues in an expert analytical brief that accounts for all associated error logs, and delivers tailored resolution steps as soon as you log in.
• Where it's available: Surfaced through proactive login notifications and detailed in the Alerts tab. The skill can also be invoked from the chat at any time.
• Why it matters: Minor implementation breakages or stale feeds often go unnoticed in background tabs until business metrics suffer. This moves your operations from manual tracking to proactive, AI-assisted alerting and remediation.
• Specifications:
  • Monitored categories:
    • Implementation: Missing pageviews, page context errors, custom event errors, and Experience API failures.
    • Data feeds: Product feed sync failures, stale user data uploads, and feed format errors.
    • Performance anomaly (coming soon): Sudden, statistically significant drops in section-wide traffic or core conversion metrics.
  • Resolution output: Delivers exactly 3 actionable investigation steps grounded in official documentation, pairs performance anomalies with 28-day trends, and generates direct download links for associated error logs.

Recommend affinity properties

• What it does: Led by the analyst agent and the personalization expert agent, this skill scans your entire active product feed to rank and recommend the top 5 product attributes to activate for your advanced behavioral affinity personalization programs.
• Where it's available: In the product feed preview screen when setting up affinity profile properties, the Recommendations strategy and report pages, or from the chat panel.
• Why it matters: Selecting wrong or highly volatile properties (like a fast-changing sale flag) can cause muddy customer segmentation. This eliminates extensive data-mapping workshops by mathematically validating which fields yield clear interest signals.
• Specifications:
  • Evaluated tiers:
    • Long-term (highest priority): Structural values that remain static (such as brand, material, style, product type).
    • Medium-term (good priority): Steady signals of consumer habits (such as color, price range, lifestyle tier, occasion).
    • Short-term (avoided): Rapidly changing values (such as on-sale, trending, new-arrival).
    • Prohibited (excluded): Granular or specific parameters with zero segmentation value (such as SKUs, absolute product URLs).
  • Output format: Returns a ranked 1-5 list detailing data quality metrics (coverage percentages and unique value counts) paired with clear strategic reasoning.
• Tips for best results:
  • Download the finalized recommendation sheet as a PDF directly from the chat panel to instantly forward to your CSM for backend allow-listing.
  • Challenge the choices by asking follow-up questions in the chat, such as "Why did you choose attribute X over attribute Y?" to review the excluded candidates.

Plan a campaign

• What it does: Led by the analyst agent and the personalization expert agent, this skill runs a short conversational flow that produces a complete, stakeholder-ready campaign plan.
  • The process: It ingests your section's full setup — applications, audiences, data feeds, goals, existing campaigns, and per-experience results — and guides you through one of three paths: Surface untapped opportunities via gap analysis, solve a specific business problem, help you shape your own idea, or optimize an underperforming campaign.
  • The output: A two-section campaign plan that includes a strategic story (problem statement, hypothesis, expected impact, success criteria) and a compact execution sketch (campaign type, targeting, variations, next steps).
• Where it's available: Accessible from anywhere in Experience OS in the chat panel. The skill is also context-aware: If invoked from inside a campaign report it defaults to optimizing that campaign. If invoked from Reconnect it biases toward triggered-message opportunities. For recommendations screens it focuses on recommendation strategies. You can always override the agent suggestions by describing what you want.
• Why it matters: Connecting a business goal to the right Dynamic Yield capabilities, structuring a clear hypothesis, and sizing expected impact takes time. This skill compresses that into a 3–5 exchange conversation grounded in your actual program data, and proactively surfaces capabilities you may not have explored yet.
• Specifications:
  • Entry paths:
    • Show me my opportunities: Analyzes your setup and surfaces 2 specific opportunities you're missing, then builds a campaign plan.
    • I have something in mind: Identifies whether you're solving a problem or shaping an idea, validates against your capabilities and data, then builds a plan.
    • Optimize an existing campaign : Name a campaign or ask which ones need attention. The agent diagnoses using your per-experience data and proposes a concrete iteration. If you're unsure which campaign to pick, it triages and ranks candidates by severity with one-phrase data-grounded reasons.
  • Campaign plan output: Renders alongside the chat in a split-pane view, exportable as PDF. The first section is the strategic story (~80%), and the second is the execution sketch (~20%) with campaign metadata, targeting, and variations.
  • Adaptive and inventory-aware: The agent re-evaluates your intent at every turn and switches paths gracefully. When presenting options from your account (audiences, KPIs, campaigns), it shows the total count, surfaces the most relevant as chips, and offers a See all escape.
  • Tips for best results:
    • Describe your business goal rather than a feature name (for example, use "recover abandoned carts on mobile" rather than "Reconnect campaign"). The agent picks the best approach.
    • Export the finished plan as PDF to share with stakeholders or include in business reviews.

Analyze recommendations program

• What it does: Led by the analyst agent, this skill generates a multi-slide executive summary of your section's recommendations program. It covers total revenue impact (direct and assisted), monthly trends, revenue share over time, year-over-year comparisons, top-performing campaigns and strategies, algorithm effectiveness by page type, and prioritized growth suggestions, all rendered as a scrollable slide report with charts, KPI cards, and narrative insights.
• Where it's available: In your Recommendations report (the skill analyzes data for whatever date range you have selected in the report's date picker). Can also be invoked from the chat panel at any time, and other skills can request a specific slide programmatically.
• Why it matters: The Recommendations report gives you granular per-campaign data tables, but extracting a high-level story for leadership reviews or QBRs still requires manual analysis and slide-building. This skill produces a ready-to-share executive report in seconds, connecting revenue trends to specific strategies and surfacing exactly where your program can grow.
• Specifications:
  • Report slides (up to 12): Title, Executive Summary (4 key findings), Total Program Impact (direct/assisted revenue, purchases, CTR), Monthly Direct Revenue & Purchases trend, Revenue Share trend, Year-over-Year Comparison, Revenue by Page Type, Revenue by Page Category, Top Campaigns, Top Strategies, Algorithm Comparison by page type, and Suggestions for Growth (up to 6 prioritized recommendations).
  • Adaptive length: The report adjusts to your selected date range. Trend slides (monthly, revenue share, year-over-year) require at least 3 months of data. Shorter ranges produce a focused 9-slide report with aggregate and breakdown slides only.
  • Single-slide mode: You can ask for a specific analysis (e.g., "Which strategies perform best?" or "Compare algorithms by page type") and the skill returns just that slide rather than the full report.
  • Sharing: The full report or any individual slide can be downloaded as a PDF for immediate distribution or inclusion in presentations.
• Tips for best results:
  • Select a longer date range (90 days or more) in the Recommendations report before invoking the skill to unlock the full 12-slide report including trend analysis and year-over-year comparison.
  • Ask follow-up questions after the report generates. For example, "Why did revenue drop in March?" or "Which algorithm should I expand?" The agent will answer using the data it already analyzed.