🧰 mr-kelly/skills
Kelly's personal AI skills workspace for daily business operations.
65 skills, including 62 App-in-Skill workflows with calm local browser UIs for review, approval, and dashboards.
English · 简体中文 · 🌐 Browse all skills on the website
Not prompts — practical skills with local apps, model builders, and review workflows where the task needs them.
Contents
- What Makes These Different
- App-in-Skill Compliance
- Quick Start
- Skills
- App UI Screenshots
- Repository Layout
This repo collects the skills Kelly uses repeatedly across email, money, content, PR review, short-drama production, music-video planning, and agent setup. It is installable as a skill/plugin bundle, but the center of gravity is not a generic public marketplace — it is a practical set of Kelly-shaped business tools.
What Makes These Different
Most skill libraries are just prompts. Kelly's skills are App-in-Skill workflows — each pairs an agent operating procedure with a local browser UI, so there is always a calm place to review before anything happens:
- The agent operates — it triages, drafts, reconciles, researches, and plans against your real accounts and exports.
- You review in a local UI — a command desk with dashboards, editable drafts, detail panes, and status filters, running on
localhostwith demo-safe data by default. - Nothing risky ships without approval — safe actions are one click; risky ones are blocked at an explicit approval boundary, and you can always hand work back to the agent with notes.
The result is agent speed with a human in the loop — not a black box.
This is not ad-hoc: the pattern follows the App-in-Skill specification — a research spec for pairing an agent skill with a local companion UI (the file handoff, the five-state review model, the data-provider seam, onboarding, and safety gates). Every skill here is an implementation of that spec.
App-in-Skill Compliance
All 60 kelly-* workflows are audited against the repository's app-in-skill-creator contract. The shared baseline is enforced in code, not left as documentation-only guidance:
- First-run onboarding — every app exposes a provider-aware setup route before live data is used. Setup points users to the provider's own secure configuration flow and never collects passwords or API keys in the browser UI.
- Deterministic handoff — every workflow ships a UI-state validator and fails loudly when the agent-to-app handoff is absent or malformed. Busabase-backed workflows also declare a fingerprinted schema manifest.
- Review safety — demo data is the default; external writes, publishing, sending, generation, and other consequential work stay behind an explicit approval or agent-task boundary.
- Maintainable frontend — large browser scripts are split into native ESM modules, entry files stay below 800 lines, and larger stylesheets use ordered CSS modules with stable cascade layers.
- Reproducible evidence — screenshot routes use deterministic demo states at the canonical desktop and phone viewports, and the GitHub Pages gallery is rebuilt from the bilingual READMEs plus skill-local assets.
The audit gate runs repository lint and type checks, validates every skill package, verifies executable launch scripts and setup routes, and exercises the App UIs at desktop and mobile sizes.
Quick Start
- Install — in Claude Code:
Or for Codex and other agents that can install skills:/plugin marketplace add mr-kelly/skills /plugin install mr-kelly-skillsnpx skills add mr-kelly/skills - Invoke a skill — e.g.
$kelly-moneyto review cashflow, or$kelly-emailto reach inbox zero. - Open the local App UI — the skill launches a review desk in your browser (demo-safe data by default) where you inspect dashboards, edit drafts, and approve or block actions.
Skills
The kelly-* skills are the everyday business tools. Helper skills such as agent-rules, app-in-skill-creator, and publish-skills support the workspace itself.
| Skill | What It Does | When To Use It | Details |
|---|---|---|---|
agent-rules |
Keeps rules and skills for Codex, Claude Code, Copilot, Kiro, Cursor, and Gemini aligned from one source of truth. It creates and verifies symlinks so agents share AGENTS.md and .agents/skills/. |
Use it when setting up a repo for multiple coding agents, checking agent rule drift, or fixing broken skill/rule symlinks. | View ↗ |
app-in-skill-creator |
Documents and scaffolds the App-in-Skill pattern: local review UI, handoff files, locks, scripts, safe approval boundaries, and optional skill-local screenshots under assets/screenshots/ only when screenshots are requested or already exist. |
Use it when building a skill that needs a browser-based review queue, approval desk, dashboard, or lightweight local workflow. | View ↗ |
publish-skills |
Publishes agent skills and MCP servers to marketplaces and registries: security-scans for private data, validates with gh skill, cuts a release, wires the Claude /plugin and Codex marketplaces, and preps the MCP Registry and curated stores. |
Use it when publishing, releasing, or listing skills, plugins, or MCP servers to skills.sh, Claude Code, Codex, or the MCP Registry. | View ↗ |
kelly-email |
Runs an AI-assisted inbox-zero workflow across configured email accounts. It triages unread mail, drafts replies, prepares cleanup actions, and uses a local UI for human approval before execution. | Use it when processing unread email, drafting support replies, archiving or marking messages read after approval, or managing email through an App-in-Skill UI. | View ↗ |
kelly-finance |
Builds and audits finance three-statement models, operating forecasts, budgets, cash runway models, SaaS/unit-economics packs, and Excel-ready finance outputs. | Use it when making 财务三表, investor projections, board finance packs, scenario cases, balance-sheet checks, working-capital schedules, capex/debt schedules, or repairing broken model links. | View ↗ |
kelly-money |
Aggregates Mercury, Stripe, Airwallex, and Creem into a local money ledger dashboard with total cashflow, account health, and account detail views. | Use it when reviewing balances, payments, payouts, fees, refunds, transfers, provider sync status, or total money movement across configured accounts. | View ↗ |
kelly-invoice-sheet |
Extracts invoices, receipts, credit notes, and statements into a spreadsheet-style local review table with field confidence, line items, approval decisions, and CSV/JSON export. | Use it for Invoice转表格, invoice OCR, receipt-to-spreadsheet, bookkeeping import prep, or a Lido-style Extract Data workflow. | View ↗ |
kelly-invest-webull |
Aggregates a personal Webull brokerage account into a local read-only portfolio dashboard: holdings, cost basis, market value, unrealized P/L, day change, and allocation by asset type. Read-only — it never places or cancels orders. | Use it when reviewing personal investments, positions, portfolio value, unrealized gains, cash, or asset allocation synced from Webull OpenAPI. | View ↗ |
kelly-family-office |
Consolidates multiple entities' and members' holdings from CSV import and manual entry into one family-office dashboard: total AUM in a base currency, plus roll-ups by entity, asset class, and institution, and performance. Read-only — it never moves money. | Use it when rolling up a family office across individuals, trusts, and companies; reviewing consolidated AUM, asset allocation, custodian exposure, or unrealized performance. | View ↗ |
kelly-family-fund |
Pools elderly parents' pensions into one steward-managed fund and books the monthly care cost and shared-family spending in a base currency, so every sibling family can see the split is fair. Read-only — it never moves money. | Use it when a family jointly supports elderly parents from a pooled pension: tracking the nursing-home cost and splitting the surplus (transport, meals, birthday gifts, gifts of obligation) transparently across sibling families. | View ↗ |
kelly-crm |
Runs a personal CRM over contacts, companies, deals, and interactions with a pipeline dashboard and an agent-drafted follow-up review queue. | Use it when tracking deals and relationships, reviewing pipeline health, or approving and editing follow-up drafts before the agent sends them through other channels. | View ↗ |
kelly-messenger |
Aggregates WhatsApp, Discord, Slack, and Telegram into one unified local inbox with full conversation transcripts and an approval-gated reply outbox. | Use it when reading messages across chat platforms in one place, drafting replies in a single composer, and approving queued outbound messages that the agent then sends via platform connectors. | View ↗ |
kelly-social |
Social command desk that both monitors and publishes (Aaron's ECHO): unified timeline, account stats, follower trends, and share-of-voice on the monitoring side, plus a content calendar, an agent-drafted post composer, short-video scripts, an approval-gated engagement inbox, and a crisis playbook — every draft passes a social-qa SHIP/FIX/BLOCK gate. | Use it when reviewing social performance and share-of-voice, planning a content calendar, approving posts and short-video scripts, or triaging mentions and replies across platforms. | View ↗ |
kelly-support |
Customer-support desk: the agent triages tickets from email, WhatsApp, web chat, forms, and WeChat, drafts replies grounded in a knowledge base, and proposes actions; you review, edit, and approve in a local UI before anything is sent, with an SLA board, CSAT tracking, and a support-qa gate that blocks refunds or commitments made without approval. | Use it when running a support inbox across channels, drafting KB-grounded replies, tracking SLA and CSAT, or approving sensitive actions like refunds and escalations. | View ↗ |
kelly-seo |
Search desk covering SEO + GEO: Google Search Console analytics (clicks/impressions/CTR/position, opportunities) plus a generative-engine-optimization side — an AI-visibility tracker across ChatGPT/Perplexity/Gemini/Claude/Copilot, a citability-optimization queue, and brand-entity / knowledge-panel readiness, gated by geo-qa. | Use it when analyzing search performance, tracking and improving how AI engines cite your brand, approving GEO content changes, or fixing knowledge-graph and entity signals. | View ↗ |
kelly-feedback |
Aggregates user feedback from every channel, clusters it into weighted feature requests, and runs a roadmap decision queue with drafted replies and changelog notes. | Use it when triaging user feedback, prioritizing feature requests, or making evidence-backed roadmap promote/decline decisions. | View ↗ |
kelly-radar |
Market-intelligence desk merging competitor signal monitoring (pricing, changelog, launch, review diffs), a research question workbench with brief approval and cited reports, and keyword/topic trend tracking. | Use it when watching competitors, commissioning deep research reports, or turning rising search and community trends into opportunity cards. | View ↗ |
kelly-ai-newsroom |
AI/news-source intelligence cockpit that separates noisy attention from buyer-trigger signals and turns platform, search, regulator, and enterprise-software movement into approved sales actions. | Use it when tracking AI news, Microsoft/Meta/OpenAI/Google/Perplexity movement, news-source changes, or buyer-intent angles that can affect product purchases. | View ↗ |
kelly-real-estate-intel |
Property-agency intelligence desk for listings, transactions, mortgage movement, district updates, competitor ads, and buyer/seller follow-up scripts. | Use it when turning real-estate market movement into owner updates, buyer outreach, listing angles, open-house talking points, or agency review batches. | View ↗ |
kelly-education-intel |
Education and enrollment intelligence desk for exam dates, admissions changes, visa news, school notices, parent questions, competitor courses, and advisor scripts. | Use it when schools, tutoring centers, or admissions services need parent FAQs, enrollment actions, course-launch angles, webinars, or sourced education memos. | View ↗ |
kelly-beauty-intel |
Beauty, wellness, and medical-aesthetic intelligence desk for competitor offers, treatment trends, safety notices, review themes, seasonal demand, and safe consultation copy. | Use it when clinics or beauty teams need campaign angles, consultation scripts, review-recovery notes, or client education that stays inside safe-claim boundaries. | View ↗ |
kelly-insurance-intel |
Insurance advisor intelligence desk for regulator updates, insurer/product movement, premium and claims news, lifecycle events, and compliant client-review prompts. | Use it when brokers or agencies need sourced meeting agendas, renewal scripts, client education notes, or needs-review checklists without making suitability claims. | View ↗ |
kelly-insure-data |
Insurance data-entry and governance cockpit backed by Busabase: Drive-node file metadata, governed QA pairs, and insurance news Base records. | Use it when importing, reviewing, and cleaning insurance files, metadata completeness, QA pairs, or market/news records before they become trusted knowledge. | View ↗ |
kelly-retail-intel |
Retail operating intelligence desk for weather, events, competitor promotions, product trends, review themes, supplier movement, merchandising, and staff briefs. | Use it when stores or consumer brands need hero-SKU picks, signage copy, replenishment checks, store briefings, or local demand actions. | View ↗ |
kelly-ecommerce-intel |
Ecommerce seller intelligence desk for marketplace policy, competitor price and listing movement, search trends, ads, review language, and SKU campaign actions. | Use it when marketplace or DTC sellers need listing edits, ad angles, bundle tests, review replies, campaign briefs, or policy-risk checks. | View ↗ |
kelly-restaurant-intel |
Restaurant-group intelligence desk for weather, events, menus, delivery movement, booking signals, review themes, shift briefs, and daily promotion actions. | Use it when restaurants, cafes, or F&B groups need hero-menu picks, service-period briefs, delivery copy, booking scripts, or guest-recovery drafts. | View ↗ |
kelly-financial-services-intel |
Financial-services intelligence desk for regulatory, macro, market, portfolio-theme, competitor, and client-question signals with reviewable education drafts. | Use it when advisors, family offices, or financial-services teams need internal briefs, client education memos, meeting agendas, or risk reminders without personalized advice. | View ↗ |
kelly-devops |
Watches the product fleet: service uptime and latency, SSL certificate and domain expiry, API key rotation, and cloud spend anomalies, with agent-proposed action cards for approval. | Use it when checking service health, catching expiring domains and certificates, reviewing cloud spend spikes, or approving renewal and rotation actions. | View ↗ |
kelly-audit |
Imports orders, invoices, and payments from business exports and audits them against each other: missing invoices, amount mismatches, overdue receivables with aging, duplicates, and unmatched payments, each with an evidence chain and drafted follow-up. | Use it when reconciling the order-invoice-payment chain, chasing receivables, or reviewing finance anomalies before month-end. | View ↗ |
kelly-tickets |
Triages complaints and requests from WeChat exports, call logs, forms, and email into classified tickets, proposes crew dispatches with SLA targets for approval, and tracks everything on a resolution board. | Use it when managing property or facilities complaints, dispatching work orders to crews, or running any intake-classify-dispatch-track workflow. | View ↗ |
kelly-homework-coach |
Child-friendly homework tutoring desk: the agent explains photographed questions, analyzes wrong answers, builds a mistake notebook, and prepares practice papers behind parent/teacher review. | Use it when elementary students need photo-based homework help, gentle step-by-step explanations, wrong-question review, or mistake-focused paper generation. | View ↗ |
kelly-lesson |
Drafts lesson plans from curriculum materials and the school template, runs a compliance checklist against school requirements, and gives the dean a review queue with teacher feedback drafts and document export. | Use it when standardizing lesson plans across teachers, checking plan compliance, or reviewing and approving teaching plans at scale. | View ↗ |
kelly-ppt-factory |
Project-based PPTX factory: turns briefs, source materials, and style samples into reviewed slide cards, generates style-consistent PowerPoint decks, and tracks render QA and exports. | Use it when producing pitch decks, sales decks, training materials, reports, proposals, courseware, or any batch of client-ready PPTX files. | View ↗ |
kelly-demo-video-factory |
Plans product demo/marketing videos end to end: hook/pain-point/storyboard drafting, codebase claim verification against the actual product, and per-shot recording progress, tracked in Busabase and rendered in a local read-only review app; hands off to post-production/HyperFrame (Remotion). | Use it when planning a product demo video, writing a storyboard, verifying a video script's product claims before recording, tracking which shots are recorded, or handing a finished storyboard to editing/Remotion. | View ↗ |
kelly-inquiry |
Aggregates WhatsApp, Instagram, Messenger, and email inquiries into a sales pipeline with a product knowledge base, quote worksheets with price guards, an approval-gated outbox, and follow-up reminders. | Use it when handling foreign-trade or DTC inquiries, drafting accurate replies and quotes from a product KB, or catching stale deals before they leak. | View ↗ |
kelly-picks |
Product-research radar for cross-border sellers: agent-swept trend candidates from BSR movers, TikTok virals, and rising queries, each with a live margin card (price, landed cost, fees, breakeven ACOS) and a competition read. | Use it when hunting products to sell, pressure-testing margins before committing, or running develop/watch/drop decisions with sourcing and listing briefs. | View ↗ |
kelly-products |
E-commerce product management desk: visual SKU catalog, pricing, inventory cover, channel status, content assets, compliance notes, lifecycle state, and approval-gated product operations. | Use it when managing product master data, inventory/reorder risks, marketplace channel status, price changes, quality holds, SKU archive decisions, or publish approvals. | View ↗ |
kelly-listing |
Listing factory: drafts platform-specific listings (Amazon title/bullets/description/backend terms/A+, Shopify, TikTok Shop, eBay) with locale variants, runs per-platform compliance checks, and exports approved copy. | Use it when writing or localizing marketplace listings, enforcing banned-word and character-limit rules, or batch-reviewing listing drafts before upload. | View ↗ |
kelly-legal-casebase-ingest |
Legal casebase intake and anonymization QA desk: the agent extracts archived judgments and awards into structured, redacted case records; reviewers approve, revise, or block before canonical ingest. | Use it when building an internal intelligent case database, processing judgment documents, reviewing anonymization, classifying/tagging cases, or running casebase quality acceptance. | View ↗ |
kelly-legal-precedent-desk |
Internal precedent research desk: the agent searches approved casebase records, prepares similar-case packs, local court-pattern notes, citations, and AI Q&A answers for reviewer approval. | Use it when lawyers need internal casebase search, similar-case matching, local court tendencies, precedent packs, or approved research exports grounded in the firm's own matters. | View ↗ |
kelly-legal-matter-strategy |
Matter-strategy desk: the agent turns new matter facts and internal precedents into issue trees, evidence maps, risk posture, negotiation options, and pleading or memo outlines for partner review. | Use it when preparing litigation, arbitration, advisory, evidence, or drafting strategy that must be reviewed by the responsible lawyer before client-facing use. | View ↗ |
kelly-legal-firm-radar |
Law-firm analytics and lawyer-profile radar over anonymized casebase metadata: practice mix, case-quality indicators, talent signals, brand proof points, and approval-gated management reports. | Use it when partners need business-layout analysis, case quality review, lawyer capability profiles, talent planning, or brand proof points from internal casebase data. | View ↗ |
kelly-clm |
Lightweight contract lifecycle desk for contract inventory, lifecycle stages, owners, obligations, renewal notices, and simple approval reminders. | Use it when managing a simple contract repository, tracking renewals or notice deadlines, assigning contract owners, or following up on obligations without doing detailed legal redlines. | View ↗ |
kelly-legal-contracts |
Legal-contract review desk for NDAs, MSAs, DPAs, and SOWs: the agent prepares clause issues, fallback language, playbook checks, and issue-list exports; legal reviews and approves in a local UI. | Use it when reviewing contracts, triaging clause risk, maintaining fallback playbooks, approving redline positions, or exporting legal issue lists without sending anything automatically. | View ↗ |
kelly-ads |
Ad-campaign command desk aggregating Amazon, Meta, TikTok, and Google ads into one board with ACOS/ROAS tracking, deterministic anomaly detection, and approval-gated adjustment cards (negative keywords, bids, budgets). | Use it when reviewing ad spend across platforms, catching zero-conversion spend and budget burnouts, or approving bid and keyword adjustments with evidence. | View ↗ |
kelly-standup |
Team standup board: the agent collects members' daily check-ins from chat channels on demand, structures them into yesterday/today/blockers cards with a team digest, and drafts approval-gated nudges for missing check-ins. | Use it when running daily standups asynchronously, seeing what everyone is working on at a glance, or tracking blockers and participation across the team. | View ↗ |
kelly-writer |
Repurposes one source idea, article, transcript, outline, or announcement into channel-ready drafts for platforms like Xiaohongshu, WeChat, newsletters, LinkedIn, X/Twitter, short video, and SEO snippets. | Use it when turning long-form source material into a multi-platform content pack with local review, edits, approvals, and export. | View ↗ |
kelly-pr-review |
Runs a GitHub PR review desk through gh CLI. It gathers review-requested pull requests, prepares review notes, uses a local UI for approval, and executes approved gh pr review actions. |
Use it when reviewing GitHub pull requests, approving/commenting/requesting changes from a local queue, or batching PR review decisions. | View ↗ |
kelly-drama |
Produces short-drama series with a local workbench for series overview, character library, relationship map, episode table, and shot sheets. Generates storyboard images with character reference cards and coordinates AI and human tasks. | Use it when planning and producing a short-drama series end-to-end: writing episode scripts, building character sheets, managing storyboard shots, and reviewing AI-generated images before use. | View ↗ |
kelly-mv |
Builds a pure-visual music video workbench: upload an MP3, write the MV concept, build a cast of on-screen characters with reference cards, and create a shot-by-shot storyboard with generated or uploaded images and draft videos. | Use it when producing a pure-visual music video — no narration or subtitles — by generating or uploading shot images and videos and assembling them over the song. | View ↗ |
kelly-digital-human |
Digital-human solution desk and multimodal demo for choosing between low-cost 2D photoreal avatar services and high-control UE/Unity 3D custom digital humans, with a local studio showing voice/text input, lip-sync video stream, vendor route latency, and launch QA. | Use it when planning an AI host, customer-service avatar, product explainer, livestream assistant, or digital-human demo; comparing services such as Silicon Intelligence, Tencent Zhiying, or ZEGO-style real-time providers; or designing a 3D UE/Unity avatar pipeline. | View ↗ |
kelly-creators |
Influencer/creator-marketing command desk built on the Discover→Plan→Activate→Measure pipeline: the agent sweeps and fit-scores creator candidates (C³ ACE), drafts outreach, briefs, and contracts, and a pre-publish gate (SHIP/FIX/BLOCK) checks FTC disclosure and claim authenticity, all reviewed in a local UI with a ROI board. | Use it when discovering and vetting creators, approving outreach and briefs, running a creator-campaign pipeline, or tracking influencer ROI and budget. | View ↗ |
kelly-campaigns |
Outbound email-marketing desk on the SEND lifecycle (Setup→Engage→Nurture→Deliver): the agent builds segments, drafts campaigns, newsletters, and sequences, and runs pre-send deliverability + subject-line QA behind an EQS quality gate (SHIP/FIX/BLOCK) before anything is scheduled or sent. | Use it when planning email campaigns, newsletters, or lifecycle sequences, checking deliverability and A/B subjects, or approving sends — distinct from kelly-email inbox triage. |
View ↗ |
kelly-launch |
Product-launch command desk on the RAMP framework (Research→Assemble→Mobilize→Prove): the agent assembles the launch checklist, drafts assets, Product Hunt / Hacker News submissions, press pitches, and the launch-day runbook, with a readiness gate scoring launch quality (LQS → SHIP/FIX/BLOCK). | Use it when planning and running a product launch: building the checklist, approving assets and channel submissions, gating launch readiness, or conducting launch day. | View ↗ |
kelly-brand |
Brand-narrative single source of truth on the TALE framework (Trace→Architect→Land→Evaluate): the agent drafts positioning, the message house, story bank, evidence-backed proof points, and vocabulary guardrails, scores narrative quality (NQS → SHIP/FIX/BLOCK), and flags cross-channel drift; you curate which drafts become canonical. | Use it when defining or auditing brand positioning and messaging, curating the canonical narrative and story bank, or catching off-brand drift across channels. | View ↗ |
kelly-revshare-simulator |
Control-panel workspace for modeling revenue-based-financing (RBF) deals: projects cash flow and cumulative repayment, computes a Cash-Flow Payout Multiple and effective annualized merchant cost, and flags risk (cap not reached, cost too high) with pure deterministic math. | Use it when underwriting a revenue-share or merchant-cash-advance deal, comparing financing scenarios side by side, or recording an approve/needs-revision/reject decision. | View ↗ |
kelly-deal-scorer |
Review-queue desk that scores candidate SME financing deals with a deterministic, fully auditable rule-based rubric (revenue stability, growth trend, category risk, principal ratio, track record) — never an LLM call. | Use it when triaging a deal-underwriting queue, checking a score breakdown, getting a suggested revenue-share rate range, or recording approve/send-back/reject decisions. | View ↗ |
kelly-portfolio-health |
Read-mostly dashboard for an RBF fund or private-credit book: total AUM, weighted repayment progress, concentration risk by category, and a watchlist of contracts with declining revenue. | Use it when reviewing portfolio health, flagging a contract for review, or checking industry concentration and at-risk exposure. | View ↗ |
kelly-lead-funnel |
Kanban control panel for a BD/sourcing team triaging merchant financing leads, with a deterministic rule-based lead-quality score and per-stage funnel conversion tracking. | Use it when reviewing a deal-sourcing pipeline, moving a lead's stage, rejecting a lead with a reason, or checking funnel conversion rates. | View ↗ |
kelly-disclosure-tracker |
Compliance/IR review workspace tracking a standardized disclosure package per financing vehicle across an origination entity, a fund-manager entity, and a listing venue, with cross-entity reconciliation flags. | Use it when reviewing disclosure checklists, vehicle readiness, or reconciliation mismatches before a filing. | View ↗ |
kelly-agent-observability |
Local dashboard visualizing a fleet of LLM agents behind a shared AI gateway: call volume, latency, error rates, cost, and trace-level chain breaks. | Use it when reviewing agent fleet health or drilling into a failed trace to see where a chain broke. | View ↗ |
kelly-agent-eval |
Review board that runs a fixed suite of mock test cases against baseline vs candidate agent versions, surfacing rubric-scored regressions before a release gate. | Use it when triaging agent-version regressions, comparing baseline vs candidate quality, or recording a release approve/block decision. | View ↗ |
kelly-agent-builder |
Low-code agent configuration and governance console: a catalog of mock agent configs with quota, approval, and ownership tracking, blocking incomplete configs from going live. | Use it when managing an agent catalog, checking quota usage, activating a draft agent, or archiving one. | View ↗ |
kelly-behavior-predict |
Dashboard over mock user-behavior funnel data with per-segment predicted-next-action heuristics and a precision/recall backtest against historical sessions. | Use it when reviewing funnel drop-off, per-segment predictions, or backtesting a rule-based recommendation heuristic. | View ↗ |
kelly-llm-gateway |
Dashboard for a shared LLM gateway's cost and model governance: spend trend, cost breakdown by service/model, canary-rollout status, and deterministic cost/error anomaly detection. | Use it when reviewing LLM gateway spend, canary rollouts, or acknowledging a cost/error anomaly. | View ↗ |
App UI Screenshots
Most Kelly skills are more than chat prompts: they ship with local browser UIs for review, approval, dashboards, planning, and handoff workflows. They are useful when the agent can prepare work, but Kelly still needs a clear place to inspect context, edit drafts, compare rows, approve safe actions, block risky ones, or send the agent back with notes.
The common pattern is a local command desk: demo-safe data, status filters, detail panes, editable recommendations, approval controls, dashboards, and local handoff records. The screenshots below show the main use cases for each App UI rather than just isolated screens.
kelly-email
kelly-money
Kelly Money is a local finance dashboard for seeing money movement across Mercury, Stripe, Airwallex, and Creem without exposing live credentials or provider data in documentation. Demo mode shows the intended operating surface: total cashflow, provider/account columns, account health, invoice matching, exception review, and drill-down detail for reconciliation.
kelly-finance
Kelly Fi
No comments yet
Be the first to share your take.