The gallery of skills and plugins for your AI agents

Discover, document, and monitor what the agent harness can do — Python libraries, tools, MCP servers, and runtime capabilities — with repeatable, comparable snapshots.

Generate clean, consistently-styled matplotlib charts (bar, line, scatter, histogram, pie) from a DataFrame or CSV with one call.

Schedule a Teams reminder that sends only when the expected person has not already posted a relevant update.

Detect greenwashing in product descriptions, marketing copy, and catalog entries against EU Directive 2024/825 and the Green Claims Directive. Returns a structured per-claim findings report with risk levels, regulation references, and recommended corrections.

In Copilot Studio, re-sending an edited file under the same name fails to deliver it — the change is made but never reaches the user. This skill gives each iteration a new version-numbered filename (report_v1.docx, report_v2.docx…) so every update actually lands in the chat as its own attachment.

Route country-specific SharePoint questions through Country metadata first, then search only the complete set of matching documents.

Redact the 18 HIPAA Safe Harbor identifiers from clinical text (or produce a Limited Data Set) with consistent pseudonym tokens and an audit manifest of what was removed.

Redlines a document based on changes from a template with Track Changes.

Convert documents between Markdown, HTML, PDF, Word, PowerPoint, Excel/CSV and text — fully offline, using only the libraries already in the agent sandbox.

A morning brief that reads your mail, calendar, and Teams, then tells you what you owe people before the week starts.

Check a PowerPoint deck, Word document, HTML page, or Markdown file against Microsoft's Accessibility Checker rules (missing alt text, untitled slides, headerless tables, low contrast, illogical reading order) and fix what it finds.

Fill an existing AcroForm PDF's real form fields from supplied data and flatten the result into a finished, non-editable PDF

Watches your Teams chats, meeting transcripts, and Outlook mail for things people actually asked you to do, and files each one as a Microsoft To Do task with a due date, priority, and source link.

Design a rigorous, platform-aware evaluation for an AI agent - define what good looks like, pick the right grading method, build a test set, and turn results into a defensible go/no-go decision.

Turn a live agent's analytics and transcripts into a diagnosis and a prioritized improvement backlog.

Adversarial assurance review for agents you own: prompt injection, oversharing, leakage and tool misuse, mapped to fixes.

Turns a customer name, line of business, and personas into a ready-to-seed Microsoft 365 Copilot demo — fictional Word/PowerPoint/Excel example files plus a full delivery and provisioning plan, with optional one-click seeding to OneDrive and Teams.

Turn any AI/agentic use case idea (chat, uploaded doc, or attached file) into an evidence-grounded, rubric-scored assessment with a customer-branded HTML report. Guides the user question-by-question through intake, categorisation, strengthening, scoring, and reporting — grounded in the Agentic Use Case Assessment Rubric v2. Runs in Microsoft Scout and Microsoft 365 Copilot Cowork.

Redesign your existing processes with AI-first thinking

Turn engagement notes into anonymised, publishable case studies without exposing client-confidential details.

Assistente de legislação trabalhista brasileira (CLT) para equipes de RH e Departamento Pessoal. Responde dúvidas sobre férias, 13º salário, aviso prévio, jornada e tipos de rescisão, e calcula verbas rescisórias com memória de cálculo completa usando as tabelas oficiais vigentes de 2026 (salário mínimo, INSS e IRRF, incluindo a Lei 15.270/2025).

A weekday-morning Scout automation that finds emails you sent that asked for something and never got an answer, reports how many are waiting, and prepares follow-up drafts in each thread's own language that it never sends.

A six-playbook toolkit for agentic B2B sales outreach: prospect research briefings, cold emails, LinkedIn/social DMs, follow-up cadences, ad copywriting, and objection handling. Configurable via a company profile template, works for any industry, market, and language.

Audit blog posts or a blog library for clarity, structure, evidence, audience fit, and improvement priority.

Restructure draft or existing blog posts into a stronger narrative without inventing new claims.

Generate a silent, looping 1920x1080 MP4 for a conference booth, kiosk, or lobby screen — rendered frame-by-frame in Python (Pillow + ffmpeg), with previews before the full render.

Ensure every generated PowerPoint deck or Word document starts from the correct bundled or SharePoint-hosted brand template.

Rewrite drafts into a configurable house style, preserving meaning while removing generic AI phrasing.

Shows where your own week has room to breathe and where it does not, from your calendar and what you send: meeting density, back-to-back runs, deep-work blocks, evenings and weekends, and the load already building three to ten days out. When a day or a coming week is overloaded it points to the most movable meetings, and it recognises time off rather than reading it as a quiet week. Reports on you only, never on anyone else.

Run evidence-driven UAT on any browser-based app — Copilot Studio, Power Platform, Dataverse, Dynamics 365, admin centres, or a custom web app — with Playwright execution, a screenshot evidence ledger, and failure classification that separates product defects from tenant config.

A practical operating system for Copilot Studio that gives agents a consistent way to clarify outcomes, choose the right approach, solve problems, make decisions, plan projects, improve processes, handle incidents, and turn repeatable work into reusable skills. Unlike a standard Copilot experience that relies mainly on the prompt and available context, Business OS adds a structured way of working so responses are more consistent, actionable, and easier to reuse across real business tasks.

Weekly HTML email digest of open conference Call for Speakers (CFP) opportunities matching your topics of interest. Pulls from confs.tech, Papercall, AdatoSystems' CFP tracker, CFP Radar, Sessionize's public user-groups directory, community-specific hubs (like communitydays.org for the Microsoft ecosystem), targeted search of standalone Sessionize conferences and run.events, and general web search. Filters to genuinely open CFPs only, and includes close dates, event dates, format (virtual/in-person), and speaker benefits when stated. Won't repeat a listing week to week unless its deadline is closing soon.

Turn a short interview into a charming, vibrant 3-slide marketing-campaign deck by filling a bundled PowerPoint template.

Play a deterministic, resumable Colossal Cave adventure with exact game output and built-in diagnostics.

Produce a polished, narrated demo video of a live web app or Copilot Studio agent — real screen-flow footage under an AI (Ava neural) voiceover — either fully headless (ffmpeg + edge-tts) or assembled in the Clipchamp web UI as an editable project.

Comments Word or PowerPoint files with Comments.

Analyzes your Teams chats and emails to build a reusable profile of your writing voice — greetings, tone, length, punctuation, sign-offs, common phrases, and quirks — and saves it to memory for other assistants to use.

Build an interactive comparison app, either a sales battlecard advocating for your product against named competitors, or a neutral, unbiased comparison across two or more peer items with no side taken. Filterable by item and category, with live search, a compact scorecard, and copy-to-clipboard talking points. Every claim is tagged verified or unverified so nothing invented ends up in front of a customer. Exports to PowerPoint on request, with support for a real branding template.

Create CFP-ready talk titles, abstracts, takeaways, speaker notes, and submission copy from a topic.

Score documents, pages, posts, or artefact libraries against a configurable quality rubric and recommend fixes.

A Monday-morning Scout automation that scans authoritative Microsoft sources for the past week's Copilot, Copilot Studio, and agent news and posts a concise, linked digest to Teams.

Using exported files from MAC usage reports, build html Copilot Adoption Dashboard with integrated framework that suggest next steps

Tells you which Microsoft 365 / Copilot Studio agent option fits your scenario — use M365 Copilot as-is, build a standard (declarative) agent, or build a custom (custom engine) agent — and which Copilot Studio harness (GitHub Copilot, standard, or Copilot chat) to build on.

Create, import, validate, authenticate, package, evaluate, and troubleshoot Microsoft Copilot Cowork plugins with repeatable engineering guardrails.

Build a validated, paste-ready Adaptive Card package for Copilot Studio standard-harness agents, with wiring, mappings, accessibility guidance, sample data, and a text fallback.

Test a Copilot Studio agent against your own Q&A set and get a graded pass/fail report — no browser automation.

Choose the right Copilot Studio harness and build shape; in Cowork, use GPT-5.6 with High effort for Detailed mode, Claude Opus 5 as fallback, or Auto for Quick mode.

Assess whether an uploaded or exported document set is ready to power a Copilot Studio agent. Produces a corpus-based readiness score, prioritized cleanup backlog, chunking and metadata guidance, and a test-prompt suite, while calling out any evidence gaps that require live system review.

Check an MCP server against the constraints Microsoft documents for Copilot Studio before you wire it up: transport, tool schema shapes that silently drop a tool, supported authentication, orchestration fit, and connector data policy. Returns a per-check verdict with the fix on the server side.

Reads an exported Copilot Studio agent and generates a graded, runnable test suite (happy-path, paraphrase, disambiguation, negative, knowledge-grounding, multilingual, and safety cases) plus a regression set, ready to run in the free Copilot Studio test panel.

Turn a one-line use case into a build-ready Microsoft Copilot Studio agent blueprint: recommended agent type and orchestration, topics, tools, knowledge, variables, a welcome Adaptive Card, security and Copilot Credits notes, and a first test plan.

Turn a Cowork conversation, into a one-page HTML use case write-up — narrative, impact figures, workflow steps, and outputs.

Search any Microsoft Dataverse table, browse matching records, and drill into full record details — sourced live via the Dataverse MCP Server.

An enterprise design review framework that helps architects build secure, scalable, governable, and production-ready Microsoft Copilot Studio agents.

Build a personalized study plan app for an exam or certification: overview timeline, day-by-day schedule with a calendar picker, domain breakdown, one-click .ics download, and a notes tab to track confidence over time.

Create narrated, captioned 1080p explainer videos with researched scripts, generated b-roll, and supplied screenshots.

Generate clean, consistently-styled matplotlib Gantt charts from a schedule CSV or DataFrame — with group colours, completion overlays, and a today line.

Audit environmental claims in CSV, XLSX, DOCX, PPTX, conversational text, and public websites with separate ANY, CA, EU, and UK findings.

One-tap way to bundle every file your Copilot Studio agent produced in a session into a single timestamped .zip and hand it back as a download.

Catch up after annual leave in one pass — a single prioritised briefing of the mail, meetings, and Teams conversations you missed, with deep links and one-click follow-up actions.

Refines an existing plan, decision, or draft through relentless, branch-by-branch questioning. Grouped in batches of 2-5 with recommended answers, pulling facts from Work IQ where possible, until every part has been fully thought through.

Clean up Outlook inbox clutter with a review-first triage plan that groups newsletters, notifications, old meeting mail, resolved threads, and duplicates - approved per bucket, moved (never deleted) to folders you can restore from.

Preserve a departing senior leader's institutional knowledge — decisions, rationale, relationships, and tribal knowledge — by mining their M365 signals into a structured, multi-phase archive a successor can ground on.

Review uploaded knowledge-source files for duplication, redundancy, staleness, overlap, and potentially conflicting guidance, then produce an evidence-based curation backlog.

Writes KQL for Sentinel, Log Analytics and Defender XDR without inventing table or column names — checks every identifier against its real source, Microsoft documentation or your own tenant's schema, and stops instead of guessing.

When a health system's lab results ingestion pipeline hits an unknown column header, this skill semantically matches the source column against the canonical clinical schema (LOINC-anchored), scoped by lab and clinical domain, and writes a suggested mapping to a Dataverse review queue for a clinical informatics steward to approve.

Create evidence-led LinkedIn posts, promos, newsletter intros, and long-form copy from supplied facts or drafts.

Create credible LinkedIn content from facts, notes or drafts, with configurable voice and no invented evidence.

An AI-powered Instructional Intelligence Skill that helps Microsoft Certified Trainers design, deliver, assess, localize, and continuously improve world-class Microsoft learning experiences across Azure, Business Applications, Data & AI, Modern Work, and Security.

Analyzes meeting content pasted as text or provided as audio/video transcripts. Delivers a structured intelligence report: explicit decisions and action items, participant persona profiles, and — most importantly — hidden insights: unspoken tensions, implicit risks, unresolved topics, and signals that were not made explicit during the meeting but are critical to the context.

Discover authorized Microsoft 365 meeting transcripts and turn them into a reviewed, searchable library of verified timestamped moments, delivered as portable HTML, CSV, and JSON files.

Guides a customer through a discovery interview, recommends the right Microsoft AI platform (M365 Copilot, Agent Builder, Copilot Studio, Foundry, Foundry Agent Service, Windows AI Foundry, or Agent 365 as the governance layer), scores technical complexity and risk, and plots them on a 2x2 quadrant chart with planning guidance.

Run Monte Carlo simulations from natural-language risk inputs — triangular, normal, uniform, or log-normal — and return percentiles, a histogram PNG, optional interactive HTML, and a downloadable results spreadsheet.

Mines your sent emails to extract your real writing voice — languages, tone modes, structure, vocabulary, taboo phrases — and generates a personal voice skill that makes every email and Teams draft sound like you.

Scans your recent Microsoft 365 signals to surface recurring patterns worth productizing — things you keep explaining (blog candidates) or multi-step tasks you keep doing by hand (automation candidates).

Twice a week, scans your recent Microsoft 365 signals and posts a Teams summary of recurring patterns worth productizing — things you keep explaining (blog candidates) and multi-step tasks you keep doing by hand (automation candidates). Read-only and privacy-aware.

Extract tables from a PDF document (e.g. contract rebate or pricing schedules) and convert them into a clean, workable Excel spreadsheet or CSV.

Pre-test an internal launch, comms, or enablement artefact against your own role-based personas before it ships — surfaces blackspots, domain gaps, credible-detractor risks, and concrete edits. Bring your own personas file.

Turn a topic or a pile of source material — a newsletter, news digest, or set of articles — into a two-host, NotebookLM-style podcast episode, with multi-voice SSML and optional Azure Text-to-Speech audio.

Generate sharp, specific point-of-view lines for posts, slides, talks, campaigns, or positioning work.

Assess Power Automate Desktop and hybrid automation projects with evidence-based findings and prioritized remediation guidance.

Turns a Power Automate solution .zip into a clean markdown reference for every flow inside it: trigger, plain-English process, connection references, and a read/write/delete table for everything it touches. Maps which flows call each other and unresolved connections automatically.

Review a Power BI or Analysis Services semantic model (TMDL, BIM/JSON, or a pasted table/measure list) for relationship, DAX, date-handling, and naming issues, with corrected DAX for anything flagged.

Creates polished PowerPoint decks from a JSON specification using python-pptx. Designed as a Copilot Studio Skill that runs natively inside the agent's Python container (no Azure Function or custom connector required). Supports 8 layouts plus native charts (bar, column, line, pie, donut with stacked / 100% stacked variants).

Writes natural, first-person spoken presenter scripts for the slides in a deck. Treats existing speaker notes as authoritative when they contradict slide text. Calibrates length to a target duration and speaking pace and returns a timing table. Never modifies visible slide content.

Turn process notes and transcripts into an editable SOP, process map, RACI, control register, improvement backlog, and polished PowerPoint briefing.

Make branded, scannable QR codes (PNG, SVG, PDF, JPG, WEBP) for a link, Wi-Fi, contact card, email, SMS, phone, map, or event, with optional shapes, colors, logo, caption, and outer frame.

Configure once, then on a schedule sweeps a locked list of authoritative sources (auto-discovered at setup, confirmed by the user) plus any user-supplied seeds. Classifies each item, flags items relevant to the user's team using a light WorkIQ-derived profile, and renders a self-contained HTML dashboard. A tightly-bounded fallback web search is used only when a locked source is silent in the window. Domain-agnostic (tax, privacy, AI/ML, healthcare, finance, ESG, labor, etc.).

Create polished, self-contained HTML slide decks with keynote-style visuals, keyboard navigation, themes, animations, and reusable presentation components.

Diagnose the true cause of a problem — not just its symptoms — using structured techniques like 5 Whys, Fishbone diagrams, and contributing factor analysis.

Visualize routes and locations on a map. Accepts pre-ordered stops with coordinates from upstream connectors (Azure Maps, Bing Maps, Dataverse, CRM) plus optional road geometry, leg distances, and durations. Produces PNG, interactive HTML (Leaflet/OSM), GeoJSON, KML, deep links, and QR codes. Fully offline Python engine — road geometry rendered directly when provided, or OSRM used browser-side as a schematic fallback.

Turns data into a scroll-driven HTML story.

Extract meaningful photos, diagrams, charts, maps, screenshots, and other visual evidence from PDFs with context, provenance, quality checks, and duplicate detection.

Generate a requested file, enforce a safe pre-upload size gate, and upload it to SharePoint without streaming bytes through the model.

Turns any SharePoint list data into a clean, consistent markdown table. Dynamic columns based on list type and query, plus a one-click Open link for every row.

Automatically discovers and validates a SharePoint list within a connected knowledge source, analyzes its structure and data, identifies key business insights, and generates a downloadable interactive HTML report. The report includes dynamic filters, interactive charts, sortable and searchable tables with pagination, detailed record drill-down through modal popups, and direct links to open items in SharePoint. If the requested list is not found, the skill suggests available lists and closest matches before stopping the analysis.

Help makers design concise, reusable Agent Skills with clear triggers, instructions, resources, and packaging.

While you're out of office, watches a group chat and answers questions from your local knowledge docs (clearly marked AI-generated), logs anything it can't answer for later, and pings you on Teams if its setup is incomplete.

A principled negotiation intelligence skill that combines Stoic decision disciplines with evidence-backed research to help individuals, executives, and analyst agents prepare, analyze, and execute high-quality negotiations. Supports interactive scoping workflows, Deep Research sourcing (citations and source URLs), and audit-ready outputs.

Turn your notes, slides, PDFs, or pasted text into source-grounded flashcards and a practice quiz — delivered as a self-contained interactive HTML study app with spaced repetition.

Populate an existing Word template from supplied data while preserving branding, repeating table rows, and reporting missing values and changes for review.

On-demand tool/action trace for Copilot Studio runs (`/special-debug tool-trace` → `tool_trace.json`).

Price a business trip from live fares in your corporate booking tool, benchmark the non-bookable lines against your own approved expense reports, and produce an estimate a manager can approve — with a hard guardrail that never enters the booking funnel.

A Scout automation installer that, only during a configured vacation window, scans work email and Teams and forwards genuinely urgent items to your personal email.

First-pass review of a vendor contract or SOW for the clauses that commonly cause problems (auto-renewal, liability caps, termination, IP ownership), flagged by risk level to prepare for legal review, not replace it.

Makes everything the agent builds or reviews conform to WCAG 2.1 AA — HTML pages, SPAs, theming, PCF controls, model-driven and canvas apps, and Power Pages.

Routes single-output requests to the best-fit included Microsoft 365 Copilot 1p agents and reserves Cowork for long-running or multi-output work.

Turn a photo of a hand-drawn whiteboard or process sketch into a polished, client-ready infographic slide (.pptx), generated natively in the agent's Python container.

Rewrite or draft prose so it avoids the tell-tale signs of AI writing, using Wikipedia's 'Signs of AI writing' field guide.

Turn four weeks of Calendar, Mail, and Teams chat activity into a colorful dashboard of reconciled Work IQ counts.