Lanoko AI

Accounting reference from basic bookkeeping to professional practice — double-entry, financial statements, reconciliation, and reporting guidance for finance agents.

Transform an agent from reactive to proactive: heartbeat + task-queue templates that keep an agent working without waiting for prompts.

Build long-running autonomous agent loops using five composable primitives: typed memory, task files, project grouping, template schemas, and heartbeat loops.

Cheat-sheet for the agent-browser CLI: headless browser automation optimized for AI agents with accessibility-tree snapshots and ref-based element selection.

Multi-agent team coordination: routing rules, task lifecycles, handoff protocols, and review gates.

Meta-agent skill for orchestrating complex tasks through autonomous sub-agents: decompose macro tasks, spawn specialized sub-agents, coordinate file-based communication, and consolidate results.

Periodic self-reflection on recent sessions: analyzes what went well and wrong, then writes concise, actionable insights to workspace files. Designed to run as a cron job.

Stand up multi-agent teams with defined roles, task lifecycles, handoff protocols, and review workflows.

Delegate-cron-check-review pattern: async sub-agent delegation with brief generation, scheduling, and escalation.

Local openpyxl/pandas toolkit that reads, writes, merges, transforms, filters, and validates Excel (.xlsx/.xls) files. 17 offline Python scripts, no network.

Playbook for designing and implementing automation workflows to save time and scale operations — identify repetitive work, map triggers/actions, choose tools.

Sync and query CalDAV calendars (iCloud, Fastmail, Nextcloud, and other CalDAV hosts) using vdirsyncer + khal.

Zero-dependency D1 database skill for Cloudflare Pages — migrations, queries, schema validation.

Systematic code-review checklist covering security, performance, maintainability, correctness, and testing with severity guidance.

Write persuasive copy for landing pages, emails, ads, and sales pages — frameworks, hooks, and voice guidance for marketing agents.

Lightweight file-based CRM for agent-driven contact management and sales pipeline tracking.

Data analysis and visualization guidance — query data, generate reports, automate spreadsheets, and turn raw data into clear insight. Self-declares no external network.

Iterative design refinement via screenshot, evaluate, plan, apply, compare loop.

Essential Docker commands and workflows for container management, image operations, and debugging.

Create, inspect, and edit Microsoft Excel workbooks and XLSX files with reliable formulas, dates, types, formatting, and recalculation guidance.

CI/CD pipeline and git workflow with branch protection, PR gates, post-merge adversarial review, and stable promotion.

Remove signs of AI-generated writing from text so it reads naturally. Tool access limited to Read/Write/Edit/Grep/Glob.

Static-site i18n — locale management, translation resolution, marketing copy formatting, HTML helpers.

Static + runtime analysis to find stubs, broken interactions, and disconnected endpoints in web apps.

Cron-driven sidecar that closes the orchestrator dispatch gap: every 5 minutes it cheap-scans the shared-artifacts task queues for `queued` tasks assigned to a specialist and drives ONE gateway pickup turn per idle specialist with work. Zero LLM cost when nothing is queued (file check only, exits immediately). Without it, a dispatched task sits `queued` for up to the 120-minute specialist heartbeat, because `openclaw system event` wakes only the DEFAULT agent.

Cron-driven sidecar that triggers the local Mission Control stale-approval sweep every 5 minutes. Without it, /api/cron/detect-stale-approvals is defined but never invoked, so pending approvals never escalate to alerts. NAT-friendly: only POSTs http://127.0.0.1:<MC_PORT>/api/cron/detect-stale-approvals over loopback.

NAT-friendly Telegram callback poller. Long-polls getUpdates and forwards every callback_query update to the local Mission Control approval-queue/telegram/callback route. Use whenever a customer wants Telegram approve/reject buttons to work behind NAT (95%+ of installs) without a public webhook tunnel.

Cron-driven sidecar that polls or-key-worker for OpenRouter usage events and forwards each into the local Mission Control ledger via /api/usage/openrouter. Closes the cache-panel data gap — MC's cache hit-rate widget needs the cache fields the proxy worker extracts. NAT-friendly: only outbound call is a GET against or-key-worker.

Convert PDF, Word, PowerPoint, Excel, HTML, images, audio, and more to Markdown using a pinned markitdown release.

Publish a skill, agent, or module to the Lanoko Marketplace on your behalf, conversationally: scaffold the artifact, fill the manifest, run the pre-publish checklist, and publish — the agent drives the whole flow so you never touch the API. Runs entirely against your local Mission Control.

Decide when to create durable Mission Control entities — projects, documents, queued tasks, approval items, agentic requests — from natural conversation.

Query 16 web search engines (7 CN + 9 global) with advanced operators, time filters, and site search. No API keys required.

Typed knowledge graph for structured agent memory and composable skills — create and query entities and relationships. Stdlib-only Python with path-traversal guards.

Turn task-following agents into proactive partners that anticipate needs and continuously improve. Ships strong built-in prompt-injection defenses. DEEP-AUDIT reviewed.

Local search/indexing CLI (BM25 + vectors + rerank) with MCP mode, backed by a pinned qmd release and local Ollama.

Review code for security vulnerabilities, audit OWASP Top 10, and implement secure auth flows. Adapted from buildwithclaude by Dave Poon (MIT).

Agent logs its own findings to .learnings/ and, gated by thresholds, promotes durable lessons into CLAUDE/AGENTS/SOUL/TOOLS.md. Opt-in hooks inject static reminder text. DEEP-AUDIT reviewed.

End-of-session protocol that flushes context to memory files, updates PARA notes, commits changes, and provides a summary before starting a new session.

Heartbeat-driven skill version checker and auto-updater via the signed marketplace index. SPEC-142 Layer 1 — silent patch + security updates only; never overwrites user-edited files.

Slack tool actions for OpenClaw's native slack tool: send/read/edit/delete messages, react, pin/unpin, list pins/reactions/emoji, member info.

Query, design, migrate, and optimize SQL databases — SQLite, PostgreSQL, MySQL — schema design, indexing, and tuning reference.

Stripe checkout, subscriptions, and webhook handlers as a zero-dependency skill.

Heartbeat-callable skill that produces a per-day rollup of an agent's last 24h conversation, stores it in memory/daily-summaries.jsonl, and pushes the same summary into Mem0 so it becomes part of the queryable long-term memory. Wave D / SPEC-Context-Optimization.

Expert frontend design guidance for building beautiful, modern UIs — landing pages, dashboards, and app interfaces.

Mobile-first UI/UX design guidance — color systems, typography, accessibility (WCAG 2.2), and Tailwind + shadcn/ui patterns.

UI/UX design intelligence: 50+ styles, 161 color palettes, 57 font pairings, 161 product types, 99 UX guidelines, 25 chart types across 10 stacks. Offline BM25 search over bundled CSVs.

Scaffold modern, animated, responsive websites with design tokens, scroll-reveal animations, and i18n support.

Cloudflare Pages CI/CD for static websites — scaffold, deploy, cache, domains, validation.

Create, inspect, and edit Microsoft Word documents and DOCX files with reliable styles, numbering, tracked changes, and tables.

Customer experience generalist. Watches support tickets and call recordings for the same problem said five different ways, then writes the fix brief for the team.

Pulls metrics from your local Stripe/QuickBooks exports, drafts the monthly investor email, and queues it for your review. Figures come straight from your exports — always verify before you send.

Plans + drafts a monthly SEO content calendar against your CMS export. Cluster mapping, keyword gaps, internal-link planning — for B2B SaaS where intent matters more than volume.

Redlines vendor contracts (PDF or Word) against your playbook, flags non-standard clauses, suggests alternate language. Provides analysis, not legal advice — a real lawyer signs off.

Sorts a year of local statements, receipts, and 1099s into IRS- or HMRC-shaped schedules. Sensitive financial data stays on your machine. Does not file returns — that’s for your accountant.

Executes your weekly ops checklist — close books, rotate keys, archive Slack, run backups, refresh dashboards — by driving the actual desktop apps. Cross-app native orchestration.

Generates on-brand graphics from your local brand kit and Figma files. Social tiles, header banners, slide templates, ad variants — designed to follow your brand system.

Design and UX generalist. PR design reviews, accessibility and contrast review against WCAG AA guidelines (heuristic, not a certified audit), micro-interaction critique, post-ship visual QA — across web, mobile, and brand surfaces.

Reviews staged git diffs locally before push. Explains intent, flags risk, suggests rewrites. Runs against your local checkout — your source isn’t uploaded to a code-review service.

Builds and tracks new-hire onboarding plans, drafts offer letters, drives Calendar + Mail to schedule the first month. Confidential employment data stays local.

Finance generalist. Monthly close, cash position, board-ready summaries, anomaly response, and runway calls. Reads your local books and turns them into plain-English answers.

Researches leads, writes personalised cold emails, and syncs the sequence into your connected HubSpot. Drafts and queues each send for your review.

Queries your local PostHog/Mixpanel/Amplitude CSV exports, finds weekly insights, ships them as a digest. Multi-GB event tables stay local — no cloud upload limits.

Engineering generalist. Triages bugs, owns the sprint backlog, runs PR review, and is on point when prod breaks. Reads your whole codebase before commenting.

Daily scheduled scrape of your named competitors, diffs against yesterday’s snapshot, summarises what moved — pricing pages, hero copy, new docs, hiring posts.

Watches a folder of receipts and invoices, categorises each line, and drops them into your Numbers/Excel ledger. Suggests tax categories for your accountant to confirm, and helps match transactions against your bank-statement exports.

Clusters interview transcripts (.txt, .vtt, .srt) into themes, each grounded in source quotes you can trace back to the transcript. Your transcripts stay on your device. Output: research brief + insights deck.

Marketing generalist. Owns positioning, writes the launch posts, briefs the campaign, and turns feature shipments into stories customers care about.

Tails local logs, watches CI runs, pings you the moment a deploy or test breaks. Persistent local process — cloud agents can’t tail your laptop the way Sven can.

Transcribes and tags local Zoom/Meet/Loom recordings, then produces win/loss themes and product-feedback rollups. Multi-GB media never leaves your laptop.