AI & ML

shenseanchen Skill

Create and tear down the throwaway git worktrees used to test community PRs for waku-agent, without leaking API keys or leaving branches behind. Use when setting up to test a PR, and ALWAYS after a PR is merged, closed, or parked.

shenseanchen Bundle

Look up any Pokemon's types, base stats, and abilities from the public PokeAPI. Use whenever the user mentions a specific Pokemon by name or asks about its stats, types, or abilities.

ciao1019 Bundle

设计、实现、评审或打磨简洁且交互精致的前端页面与组件。提供项目画像、视觉判断、精选组件检索、官方 skills/MCP 与源码发现、多库集成和证据验收。适用于新页面、后台、AI 对话、文档工作台、参考还原和局部 UI 改善；纯后端、仅修复业务逻辑或机械改字时不使用。

ai-automation-tools Bundle

Change what Project Hub's scanner walks, indexes or serves — the filter chain in `Hub/hub.mjs` (SKIP_DIRS / DOT_OK / DOC_FILE / SECRET_DIRS), the node `kind` vocabulary, the `BUCKETS` and `USER_RUNTIMES` tables behind the CLI pages, and the second deny gate in `resolveId()` that every `/api/*` path check goes through. Use for "a folder is missing from the tree", "my new agent runtime doesn't show up", "add a file type to the scan", "add a node kind", "the scan is slow / the payload is huge", "the page won't refresh after I edit a file", "repoScope / hub.config.json", "a repo is missing from the overview table", and anything touching scanTree, scan(), the watcher, /api/scan, /api/file, /api/stat or /api/health.

ai-automation-tools Bundle

Work on Agent-Chat's push half — delivery sinks (folder / webhook / command), the Notifications settings tab, the opt-in list, and the stall watchdog. Covers config/delivery.json's shape, the four events (started / result / complete / stalled), the scope + opt-in rule, the four call sites that end a conversation, and why a sink may never raise. Use for "set up notifications", "get a push when a debate finishes", "ntfy / Discord / Slack / Gotify webhook", "deliver the transcript to a folder", "run a command after a conversation", "add a delivery event", "the watchdog isn't firing", "delivered folder doesn't match the zip", "my hand-written sink got overwritten", or any edit to src/orchestrator/delivery.py, watchdog.py, or src/web/api/notifications.py / delivery_settings.py.

ai-automation-tools Bundle

Design, register and verify a Windows Task Scheduler job by picking one of four proven archetypes — an unattended AI-agent CLI run, a wrapped maintenance script, a service watchdog or lifecycle task, or a plain command — then tracking it in Cronsole. Ships a runnable wrapper and registrar. Use whenever the user wants a new scheduled job, cron job, recurring task, nightly or weekly automation, watchdog, healthcheck, drive cleanup, doc refresh, or "run Claude Code on a schedule" on Windows; whenever an existing scheduled task needs re-scheduling, parking or debugging; and whenever a scheduled task reports success while doing nothing, stops notifying, or "ran" with no effect.

ai-automation-tools Bundle

Add or change a conversation format in the Agent-Chat repo — a new conv_type (debate / podcast / collaborate / yours), a sub-type preset, a seat role, an extra role like the skeptic, or a type that produces a deliverable. Decides first whether the thing asked for is a TYPE or merely a PRESET, then walks the full multi-file ritual: CONV_TYPES, _ROLE_BRIEFS, the runtime skill, the guide, the tests. Use for "add a new conversation type", "add an interview/standup/retro mode", "add a preset", "add a role to a conversation", "make this type produce an artifact", "the lead is getting its own seat and shouldn't", "change who speaks first", "seat counts are wrong", "signal='result' never lands", or any edit to src/orchestrator/conv_types.py or src/presets.py.

ai-automation-tools Bundle

Diagnose a live Agent-Chat conversation that is stuck, silent, ending early, or behaving wrong — a run that will not advance, a seat that is never handed a turn, "not your turn" rejections, a room that closed before the deliverable landed, an agent window that never joined, a database that is locked, or a conversation missing from the web UI or the hosted mirror. Works from the DB outward: read the row, read the rotation, then look at a CLI window. Use for "the debate is stuck", "conversation 57 isn't moving", "the agent isn't taking its turn", "why did this end at max_turns", "it says it's not my turn", "nothing shows in the UI", "database is locked", "the watchdog fired", or any Agent-Chat run that is not doing what it should.

ai-automation-tools Bundle

Decide whether an Edge-Radar betting strategy, sport, parameter or claimed edge is actually supported by outcomes — pick the right tool in scripts/backtest/ for the question, report the model-vs-market Brier pair rather than a bare "Brier", collect zero-risk shadow rows when a freeze or gate blocks its own evidence, and say what the sample can and cannot carry. Use for "is this strategy working", "should we unfreeze X", "is the edge real", "what does the settled book say about Y", "fit this parameter", "before/after a risk-gate change", "how many bets until we know", any claim about ROI, win rate, calibration, CLV or correlation, and any proposal to raise sizing or loosen a gate.

orderful Bundle

Author, inspect, test, and save the SuiteAnalytics Datasets the Orderful SuiteApp consumes — Packaging Data Sources for 856 ASNs and carton/pallet label datasets. NetSuite has no REST API for datasets, so this skill drives the SuiteApp's dataset actions on the agent RESTlets through a describe → edit spec → dry-run → save loop, with validation against the SuiteApp's own packaging and label column contracts. Use when the user says "build a label dataset", "the packaging dataset is wrong", "labels aren't generating for <retailer>", "describe the carton dataset", "add <field> to the label dataset", "test the dataset for fulfillment X", or "/manage-datasets".

davidgfolch Skill

Dependabot Agent Instructions

jialuohu Bundle

Use the guarded Overleaf MCP for configured projects and authorized file operations, or Overleaf Tools status. Do not bypass unavailable operations with browser or raw Git.

orderful Bundle

Reprocess a single inbound Orderful Transaction in a NetSuite customer's account by calling the SuiteApp's agent-write RESTlet. Use when a previously-failed (or stale) inbound transaction needs another run through the inbound MapReduce, or the user says things like "/reprocess-transaction", "reprocess <id> for <customer>", "retry transaction <id>", "rerun the inbound MR for <id>", or "reprocess this 850".

jialuohu Bundle

Sync published codex-toolbox, or check installed plugins, MCP servers, and skills for health and actionable setup problems. Health-only checks do not update or repair the machine.

relaticle Skill

Develops real-time broadcasting with Laravel Echo. Activates when setting up broadcasting (Reverb, Pusher, Ably); creating ShouldBroadcast events; defining broadcast channels (public, private, presence, encrypted); authorizing channels; configuring Echo; listening for events; implementing client events (whisper); setting up model broadcasting; broadcasting notifications; or when the user mentions broadcasting, Echo, WebSockets, real-time events, Reverb, or presence channels.

jialuohu Bundle

Check upstream updates or upgrade compatible toolbox plugins, skills, MCP dependencies, and managed marketplaces. Publishing requires a separate explicit $ship-toolbox invocation.

jialuohu Bundle

Maintain Research/LLM Wiki in Obsidian: ingest sources, answer source-backed queries, synthesize research, and audit existing wiki content.

relaticle Skill

Use whenever driving agent-browser against the local Relaticle app (relaticle.test and its panels) for testing, QA, business review, or UI automation. Covers Filament v5 + Livewire v4 quirks specific to this codebase: panel URL derivation (domain-routed vs path-routed, never assumed), login flows for the app and sysadmin panels, seeded credentials, Select/date-picker interaction, the $wire.mountAction gold pattern, tenant switching, Reverb/queue hazards, and session isolation. Every hard fact here is a DATED CACHED HINT. When one fails, re-derive from the running app and update this file (self-heal). Not for other sites or generic browser automation.

relaticle Skill

Mandatory point-of-use sequence for capturing annotated screenshots that go into deliverables (ClickUp reviews, bug repros, internal evidence, end-user docs). Invoke this BEFORE every screenshot capture, not once per session, so the rules are fresh in context at the moment you actually shoot. Covers the annotate → verify-crop → shoot → read-back flow, the helper JS files (annotate.js, verify-crop.js), and the audience-specific annotation rules (label vs. no label). If you are about to type `agent-browser screenshot file.png` for any reason other than throwaway debugging, invoke this skill first.

duodealteam Bundle

Duodeal REST API reference (api.duodeal.app) — X-API-KEY auth, payload conventions, inventory of the ~80 operations, filters, pagination, recurring errors. Use this skill as the API reference whenever you touch the Duodeal API or the duodeal MCP tools (POST/PUT/GET, debugging a 4xx/5xx, custom fields, media, logos, templates, filters), before generating, editing or delivering a Duodeal quote or selling page, and whenever deals, quotations, quotes, V2 blocks or the Hot Deal Score come up on the API side.

duodealteam Skill

End-to-end flow for building a Duodeal quote from A to Z with the duodeal MCP tools — tenant setup, customer, deal + quotation, lines (sections, discounts, options), branding, T&Cs / legal notice, links to deliver. Use whenever you need to create, duplicate or deliver a Duodeal quote or selling page, start from a template, clone a deal, add a second quotation, or set up a product catalog.

gabuldev Skill

Use when creating or auditing the layers of a Flutter feature or package - where a use case, repository, datasource, entity or model belongs; whether domain may import Flutter; whether a dependency between layers is inverted. Triggers on "where does this go?" in a feature, or a request for an architecture review.

nik-ti Bundle

INVOKE FIRST for any task building, debugging, or extending a LangChain, LangGraph, or Deep Agents project (Python or TypeScript) — including agents that call an OpenRouter-served or other non-Anthropic/OpenAI model through LangChain. Covers framework selection, quickstarts, create_agent/@tool patterns, custom graphs, persistence, human-in-the-loop, RAG, evals, and LangSmith.

nik-ti Skill

Orchestrate parallel browser automation using multiple Chrome DevTools MCP instances. Use when a task requires doing the same browser action across many targets simultaneously (e.g., submitting contact forms, filling applications, scraping pages that need JS rendering). Spin up 1-5 parallel Chrome agents.

nik-ti Bundle

Iteratively inspect an agent repository and optional user-provided traces, interview the user, and create, run, and audit Harbor evals one at a time. Use for agent evals, Harbor tasks, benchmark cases, verifier design, or controlled agent environments.

blauwtje Skill

Use when a decision about logic, state or data flow needs running code before it is made: a state model, an algorithm choice, or an untried integration. Fires on words like prototype, throwaway or proof of concept, or on the thought of building the real thing to find out. Not for a decided change, which implementing-batch builds, or a look, which designing owns.

blauwtje Bundle

Use when creating, editing or sizing a skill or agent, before the edit and before it is called done. Not for CLAUDE.md, a rule or an output style.

belentani7 Skill

Fan-out headless agent runs across claude/gemini/grok (and other adapters) via meta-cli, then collect results into memory/raw. Use for cross-provider review/diversity, parallel research, or same-provider worker batches — not for every task. The standing multi-provider surface.

ai-automation-tools Skill

Decide how much machinery a request deserves before starting it — answer inline, delegate to one subagent, recon-then-plan, or fan out across agents — and pick a model tier to match, so a two-line answer doesn't cost a five-agent workflow. Use when a request's size or blast radius is unclear, before spawning any subagent or workflow, when choosing which model an agent should run on, or when the user asks "how should we approach this", "is this a big job", "what's the best way to tackle this", "route this", "delegate this", "who should do this", or "which model should handle this". Also use when a task spans several repos, domains, or unknown files, and before any multi-agent fan-out to confirm the fan-out is actually warranted. The cost policy is a parameter — set it once and the routing table follows it.

belentani7 Skill

Migrate this app's LLM calls onto a Morph open model (GLM-5.2), either fully or as a scored 5% production trial. Use when the user wants to move a provider/model to Morph, A/B a Morph model against their incumbent, or run a canary on real traffic.

belentani7 Skill

EasyDiffusion - 1-click AI image generator with browser UI. Easy Stable Diffusion setup for beginners.

belentani7 Skill

HeartMuLa - Open-source music foundation model family by Fudan University. Music generation with lyrics.

belentani7 Bundle

Use when writing, reviewing, or committing code to enforce Karpathy's 4 coding principles — surface assumptions before coding, keep it simple, make surgical changes, define verifiable goals. Triggers on "review my diff", "check complexity", "am I overcomplicating this", "karpathy check", "before I commit", or any code quality concern where the LLM might be overcoding.

belentani7 Skill

Construccion de backend FastAPI con Qwen-Agent y ruteo de modelos.

belentani7 Bundle

Writes, refactors, and evaluates prompts for LLMs — generating optimized prompt templates, structured output schemas, evaluation rubrics, and test suites. Use when designing prompts for new LLM applications, refactoring existing prompts for better accuracy or token efficiency, implementing chain-of-thought or few-shot learning, creating system prompts with personas and guardrails, building JSON/function-calling schemas, or developing prompt evaluation frameworks to measure and improve model performance.

belentani7 Skill

Creacion de bases de datos de conocimiento locales con embedding para samples.