starchild-ai-agent/official-skills
31 skills106k installs403 starsGitHub
Install
npx skills add https://github.com/starchild-ai-agent/official-skillsSkills in this repo
1CoinglassCoinglass is an official Starchild agent extension that exposes crypto derivatives intelligence as registered tools for YAML-configured agents. Solo builders automating research, dashboards, or trading-adjacent assistants can pull funding rates and open interest from V2 endpoints, analyze long/short and account positioning ratios, inspect liquidation streams, and query V4 futures metadata including supported venues, aggregate coin metrics, pair-level stats, and OHLC history. The extension follows a standard register(api) pattern: tools load from a local tools directory and return names for the agent registry. You enable it by supplying COINGLASS_API_KEY and listing the tools in agents.yaml. It fits builders who already run the Starchild agent stack and want derivatives context inside the agent loop—not a standalone CLI product. Expect network calls on every tool invocation and API quota limits governed by Coinglass.10kinstalls2HyperliquidHyperliquid is a Starchild agent skill for trading perpetual futures and spot on Hyperliquid’s fully on-chain DEX. Solo builders experimenting with agent-managed trading wire the skill when they already run Starchild’s wallet service and want human-in-the-loop or scripted flows instead of hand-rolling REST and signing. Before any trade, you load wallet-policy and approve a standard wildcard policy that blocks key export while permitting the operations this skill needs across deposits, EIP-712 order signing, and withdrawals. The readme stresses choosing agent tool mode—calling hl_market, hl_order, hl_spot_order, and related tools inside the runtime—for conversational and task scripts, versus separate service scripts when your workflow lives outside chat. Metadata requires WALLET_SERVICE_URL; tools cover market data through order lifecycle and treasury movements. Treat it as advanced integration work with real funds at risk, not a backtest or paper-trading tutorial.6.7kinstalls3WalletWallet is a Starchild official skill that exposes multi-chain wallet operations—EVM and Solana—for agent frameworks. Solo builders experimenting with crypto-native agents or payment-adjacent automation install it to give an agent structured tools for reading balances, transferring assets, signing transactions and typed data, listing history, and managing wallet policies. The implementation registers a suite of tool classes (info, balance, transfer, sign, transactions, policy propose/get) through a single register function, delegating heavy lifting to shared wallet modules. Treat it as Build-phase integration work: you are connecting agent runtime to keys and RPC endpoints, not running growth or SEO. Policy propose/get hooks imply governance-style limits—review carefully before mainnet. Skip if your product has no on-chain surface; prioritize security review because signing and transfers touch secrets and irreversible chain actions.6.7kinstalls4TwitterThe Twitter skill is a Starchild official extension that exposes read-only X (Twitter) capabilities through twitterapi.io. Solo builders assembling research agents, social listening workflows, or launch analytics can register thirteen tools: advanced tweet search, tweet and article fetch, full thread context, quote tweets, trends, user profiles and timelines, follower and following lists, replies, retweeters, and user search. It is not a posting or moderation stack—data ingestion only—so you pair it with your own summarization, alerting, or content skills. Configuration centers on TWITTER_API_KEY in the environment and the extension’s register hook against the host ExtensionApi. Use during Build when your product or internal agent needs programmatic access to public X data without hand-rolling HTTP clients and response shapes.6.6kinstalls5CoingeckoCoinGecko is an agent extension skill that packages CoinGecko Pro crypto market data as registrable agent tools. Solo and indie builders shipping trading dashboards, portfolio agents, or research bots install it when they need reliable price, chart, OHLC, trending, exchange, and NFT endpoints inside Claude Code, Cursor, or similar agents instead of scattering ad-hoc fetch scripts. The extension auto-loads through ExtensionLoader and expects a COINGECKO_API_KEY in the environment, which keeps secrets out of prompts while still giving agents structured tool calls. It fits the Build → integrations moment: you already have an agent project and want catalog-quality market data, discovery lists, and global stats without maintaining your own CoinGecko client layer. Use it when your product shape is an agent or API-backed finance feature; skip it if you only need occasional spot checks in a browser.6.5kinstalls6TwelvedataTwelve Data is an official Starchild-style extension skill that connects your coding agent to Twelve Data Pro for equities and forex. It packages Python tools for real-time and batch quotes, historical time series, end-of-day bars, and reference datasets including exchanges and pair lists, plus search to resolve symbols. Solo builders use it when building finance copilots, trading journals, or market-aware automations without hand-rolling REST clients. Setup is integration-first: set TWELVEDATA_API_KEY, load the extension, and call register(api) so the host exposes the tool surface. Intermediate complexity due to API quotas, symbol conventions, and batching. Not a substitute for compliance review or execution infrastructure—it's market data access for agent workflows.6.4kinstalls7Skill Creatorskill-creator is a meta agent skill from Starchild’s official skills repo for bootstrapping production-ready skill folders instead of hand-rolling SKILL.md from scratch. Solo builders shipping with Claude Code, Cursor, or Codex use it when they want repeatable agent capabilities—checklists, integrations, or workflows—and need consistent frontmatter, optional resource bundles, and a clear body outline. The bundled initializer writes a templated SKILL.md plus folder layout under a path you choose, prompting you to delete the “pick a pattern” section once you commit to workflow-, task-, reference-, or capabilities-based organization. It fits early in a build when you are curating a personal or team skill library, and again in Operate when you iterate on prompts after production learnings. It is procedural knowledge, not a hosted API: you run the script locally, edit TODOs in direct voice, and register the folder in your agent’s skills path. Expect intermediate familiarity with repo layout and skill discovery conventions.6.3kinstalls8Skillmarketplaceskillmarketplace is a Starchild agent skill that centralizes how solo builders find and attach procedural skills to their coding agents. Instead of curling GitHub or copying SKILL.md by hand, you invoke search_skills so the agent checks what is already installed, scans the community registry, and searches the broader skills.sh ecosystem (OpenClaw, Vercel, Anthropic, and peers) before optionally auto-installing the top match. That makes it practical whenever you realize you need a capability—deployment helpers, trading utilities, k8s recipes—mid-journey without leaving the agent thread. Publishing and sharing custom skills back to the registry fits the same loop for indie builders who package repeatable workflows. The skill is user-invocable and optimized for day-to-day agent ergonomics: install once, call search_skills with a plain-language query, and continue the task with the new skill loaded. It pairs naturally with meta skills like skill-creator when you want to ship your own listing after validating behavior locally.6kinstalls9Orderly Onboardingorderly-onboarding is Starchild’s agent onboarding skill for Orderly Network, omnichain orderbook infrastructure for perpetual futures liquidity used by decentralized exchanges and custom trading front ends. Orderly does not ship a consumer app; it provides the shared central limit order book, settlement, and liquidity layer while builders own UX and distribution. The skill activates on first exposure to Orderly, when setting up AI agent tools for Orderly development, when you need to understand offerings end to end, or when you must pick the right follow-on skill or MCP capability for a task. Key ideas—omnichain CLOB from major EVM chains and Solana, self-custody with on-chain settlement, and one book serving many front ends—are framed so a solo builder can decide whether Orderly fits their perp or DEX product. It routes you toward deeper integration work rather than replacing API docs. Use it at the start of a Web3 build before you commit architecture or duplicate research across chat threads.5.7kinstalls10ChartingCharting is a Starchild agent skill for solo builders and trading-adjacent agents who want professional candlestick visuals instead of raw OHLC dumps in chat. It is invoked when someone asks for a chart, price visualization, or technical analysis plot. The skill’s central constraint is procedural: chart scripts fetch data internally, so the agent must never invoke separate market-data tools that can add tens of kilobytes of redundant context. You follow a fixed four-step loop—load a script template from skills/charting/scripts/, write to scripts/, execute with bash, then read the output PNG and surface it with markdown image syntax. Install expects pip packages mplfinance, pandas, and numpy plus a CoinGecko API key in the environment. It is a generator for filesystem artifacts, not a live trading or execution system.4.4kinstalls11Browser Previewbrowser-preview is a Starchild agent skill that bridges preview_serve and what builders see in the Preview tab. Solo and indie developers using Claude Code or similar agents inside Starchild often hit white screens, missing tabs, or asset 404s because they treat preview URLs like local dev servers. The skill defines how the right-side Preview iframe loads proxied URLs, how each serve call opens a tab, and how user-closed tabs differ from stopped backend services. It is the operational reference for registry, ports, paths, and blank-screen diagnostics when listing or fixing running preview services. Use it whenever a preview is broken or missing—not as a substitute for preview_serve itself, but as the playbook after a URL is returned.4.3kinstalls12CoderCoder is Starchild’s implementation specialist for agents that must ship working code, not placeholders. Solo builders use it when they want the model to open real files, apply targeted edits, run bash checks, and show stdout or test results as proof. The skill encodes a disciplined loop: understand context, change the smallest necessary surface, verify with python scripts or pytest, and iterate on failures before handing off. It explicitly favors edit_file over wholesale rewrites and bans guessing file contents. That makes it a strong default during Build for features and refactors, and during Ship when bugs block release. It is not a planning or architecture skill—pair it with spec or review skills when you need design before code or a merge gate after. For Prism’s audience, Coder is the hands-on builder persona inside Claude Code, Cursor, or Codex workflows where filesystem and shell access are already enabled.4.1kinstalls13Wallet PolicyWallet Policy is a Starchild agent skill that translates wallet security intent—transfer caps, destination allowlists, restricted methods, time windows—into Privy-compatible policy rule objects, then pushes them through wallet_propose_policy so the user confirms in the UI. Solo builders embedding Privy in a dapp, agent wallet, or treasury flow use it during integration work rather than hand-authoring opaque JSON. The SKILL.md is explicit: after generating rules, you must call wallet_propose_policy with chain_type, title, description, and a rules array using field_source, operator, and ALLOW/DENY-style actions. It fits Prism’s Build → integrations shelf for Web3 products and aligns with Security facets because policies gate funds movement. Version 1.1.0 metadata marks it user-invocable with model invocation enabled, so Cursor or Claude Code can run it when someone says they want to lock down wallet sends. It is not a standalone MCP server—it is a generator plus Privy workflow inside your agent session.4.1kinstalls14Preview DevPreview Dev is a Starchild agent skill for solo and indie builders who want to move from idea to something you can click in a browser without guessing whether the server actually works. It targets frontend and fullstack web work—React, Vue, Vite, plain HTML, Express, FastAPI, and similar stacks—where the payoff is a live preview in the Browser panel plus optional public community publish. Every run follows a hard gate after `preview_serve`: inspect `health_check`, remediate known failure modes (missing index, escaping bugs, blank SPA shells, bad command/port), and only then declare preview ready. That turns agent-assisted UI work into a repeatable ship loop instead of chat-only code dumps. Use it when you are prototyping a landing page, building a dashboard, or wiring a small API plus UI and need confidence the preview is healthy before you share or iterate.4kinstalls15Project BuilderProject Builder is a detailed reference skill in the Starchild official stack for turning user intents—especially recurring monitors, alerts, and reports—into verified Python task folders. Solo builders using Starchild agents start with Pattern A: prove data fetching with `core.skill_tools` auto-discovery of `skills/*/exports.py`, print JSON from `tasks/{id}/run.py`, then layer validation before scheduling. The readme emphasizes no skipping steps: Step 1 must return live CoinGecko or CoinGlass-style payloads (or Twelve Data fallback) before cron or notifications attach. It targets indie operators who want coding agents to behave like a cautious SRE, not a demo script. Complexity is intermediate because you need environment keys, Python 3, and familiarity with skill namespaces. Use during Build agent-tooling; later Ship testing should re-run the same verification commands.3.7kinstalls16ComposioComposio is a Starchild agent skill that routes all external-app operations through the Composio Gateway on Fly’s internal network. Solo builders use it when the product needs real actions in connected SaaS tools—send email, file issues, create Notion pages, or schedule calendar events—without embedding Composio SDK credentials in the agent runtime. The documented architecture keeps API keys on the gateway, uses plain HTTP to flycast, and relies on automatic user resolution from the deployment environment. You search compact tool slugs for a task, then invoke the gateway’s execute endpoints rather than calling Composio Cloud directly from application code. It fits indie agents and small teams shipping automation alongside a hosted Starchild stack, and it stays narrow: it does not replace your app’s OAuth UX or custom integration code for non-Composio APIs.3.1kinstalls17Slide Creatorslide-creator provides a style-neutral HTML and CSS foundation for 1280 by 720 presentation decks aimed at agent-assisted slide generation. It defines the core .slide container, typography scaffolding, header and tag patterns, card grids, and non-negotiable print rules so PDF or print export stays pixel-stable. Solo builders use it when they need deck structure fast—pitch nights, launch webinars, growth content repurposing—without baking brand colors into the skill itself; you copy the base into styles.css and layer your theme. The readme fragment is essentially the canonical structural stylesheet, which makes the skill a template more than a full creative director. It fits indie operators who ship in Claude Code or Cursor and want repeatable slide HTML instead of one-off Google Slides fiddling. Pair with your own copy and data; the skill does not invent narrative—it keeps layout and pagination trustworthy for downstream rendering or print pipelines.2.6kinstalls18ChartChart is a Starchild official skill for producing interactive, shareable visualizations when solo builders need more than a static image. It standardizes one folder per chart project under output/chart-html/, including index.html, a reproducible generate.py script, data.json snapshot, README notes, and screenshot.png. Rendering uses Apache ECharts from CDN; PNG capture relies on native getDataURL with canvas merge for dependable exports. Triggers cover BI-style asks—BTC versus gold, revenue by quarter, return series comparisons—and the workflow starts from template selection or custom layout. Version 3.0.2 emphasizes project-based iteration instead of a global gallery, which fits indie demos embedded in landing pages, investor updates, or internal dashboards. Agents can regenerate charts from saved data and scripts, making the skill a generator front-end on top of whatever analytics query or CSV you already have.2.3kinstalls19Web Crawlerweb-crawler is a Starchild official agent skill that combines web scraping with social-platform data extraction for solo builders doing research, content repurposing, or competitive listening. It activates when normal fetching fails on JavaScript-heavy sites or when you need YouTube transcripts, TikTok comments, profile posts, trending clips, or engagement-style metrics from major networks listed in the skill tags. The skill routes each request to an appropriate backend so you describe the outcome—not the vendor. It expects Python on the PATH and discourages wasteful paid API calls when a simple fetch suffices. Typical workflows span idea-phase audience research, validate-phase proof from real discussions, and grow-phase content mining, always on public data with deliberate scope.2.1kinstalls20User Onboardinguser-onboarding packages reference notes from Starchild’s five-minute guide so agents can coach new users without re-fetching the marketing page. It positions Starchild as a practical outcome-first assistant and walks through smart model routing, sharper prompts, optional delivery channels, external connectors, and quick build-preview loops. Solo builders installing Starchild skills benefit when their agent explains how to get reliable results on day one—cheaper routes for simple tasks, stronger models for hard problems, and fewer round trips because context and format are stated up front. The skill is editorial and meta: it does not execute integrations itself but steers setup order and communication style. Use it whenever someone asks how to start, waste less spend, or wire notifications and tools before attempting ambitious automations.1.7kinstalls21Chatgpt Codex OnboardingChatGPT Codex Onboarding is a protected, script-delivery skill that walks your agent through connecting an existing ChatGPT or Codex subscription using OAuth device-code login instead of a vendor API key. Solo builders install it when they pay for Plus, Pro, Team, or Enterprise and want gpt-5-codex-class models inside their coding workflow without managing BYOK secrets. The file is explicit about false triggers: similar vendor names or a vague “add OpenAI” request should prompt a clarification between subscription OAuth and API-key setup, which belongs to the separate byok-custom-model skill. Execution is bash-oriented exports from SKILL.md rather than registered tools. Prism catalogs it under agent tooling because it is a one-time (or occasional) connectivity step at the start of a build session, not a product feature you ship to customers.1.3kinstalls22Byok Custom ModelBYOK Custom Model is a script-mode agent skill for solo builders who want their coding agent on a specific LLM—cheaper OpenRouter route, company Anthropic key, self-hosted endpoint, or niche vendor—without fighting manual JSON config. Version 2.1.0 documents an API-example-first flow: you paste the provider’s sample request, the skill parses base URL, model id, and openai versus anthropic wiring, then you confirm and call add or add_template. It is distinct from ChatGPT Codex OAuth onboarding and pairs with broader model-onboarding references in the Starchild stack. Use it when default platform models are wrong for cost, geography, capability, or compliance. Expect bash-driven setup, protected delivery, and responsibility for key rotation and spend limits yourself because traffic goes direct to the vendor.1.3kinstalls23Community PublishCommunity Publish is a Starchild agent skill (v0.13.1, script delivery) that gives solo builders one sanctioned path when they want to publish, share, list, distribute, deploy publicly, or open-source something they already built. It separates three meanings of “share”: point-to-point public URLs for live previews, gallery discovery on the Project Dashboard, and pushing source to the community GitHub repository. Triggers match natural user intent—"make this public," "open source my project," "put it on the dashboard"—and the metadata tells agents not to invent bespoke deploy scripts or hunt the filesystem for half-finished publish utilities. Use it at Launch when distribution matters, and again in Operate when you refresh a preview or re-list after iterations. It complements Build skills that create the app; this skill is how the outside world sees it.1.1kinstalls24Xai Grok OnboardingxAI Grok Onboarding is a script-delivery agent skill that signs you into xAI with standard device-code OAuth so subscribed accounts can drive Grok-family models without managing a separate xAI API key. Solo builders stitching multi-agent workflows need a repeatable login path that respects subscription tier: the skill documents observed tier-to-model mapping and points operators to status() after login. Use when you want SuperGrok, X Premium login, or grok-4.x models in Starchild-style setups; pair mentally with byok-custom-model when you prefer vendor API keys instead. It is integration-focused—read the protected SKILL, run bash exports, and treat credentials as secrets. Prism lists it under build agent-tooling because provider onboarding is infrastructure for the coding agent, not a ship-time security audit or launch copy task.1.1kinstalls25VideoThe Starchild video skill is an agent skill for end-to-end generative video through fal.ai on the Starchild paid proxy. Solo builders and indie teams use it when they want cinematic or product clips without maintaining queue polling, cost accounting, or download paths themselves. Invoke it for any video-generation request on Starchild: pick a model tier, duration, and prompt (or reference media), then run the bundled Python scripts rather than calling fal directly. The workflow emphasizes operational safety—results land under `output/videos/` with cost metadata, while raw fal CDN links are unsuitable for end users because of restrictive CSP. It fits agent workflows that pair copy or storyboards from earlier phases with rendered MP4 deliverables for landing pages, social launch clips, or in-app previews.970installs26Tg Bot Bindingtg-bot-binding is a Starchild official skill that walks solo builders through creating a Telegram bot, pasting the token into the Starchild dashboard, and completing ownership verification so AI agents can talk to users in Telegram. It targets builders who want mobile-first alerts, async Q&A, or customer-facing bots without building custom Telegram middleware. Invoke it when you hear add my TG bot, bot binding code, telegram not pushing, or whitelist a TG username. The guide mirrors BotFather commands, explains token hygiene, and covers fix-it scenarios when delivery fails. For indie operators shipping one agent product, this collapses days of scattered docs into a single procedural runbook your coding agent can follow while you keep the token out of chat logs.953installs27Agentxagentx documents how solo builders use Starchild’s built-in AgentX community forum tool to share updates, ask questions, and engage peers without confusing the platform with Twitter/X. The skill must load before the first `agentx` invocation so the agent knows `create_post`, `create_thread_post`, comment trees, likes, follows, and media attachments follow the documented parameter shapes. It also encodes a critical routing rule: requests to tweet or post on X should flow to Composio’s Twitter skill, while Starchild forum language routes here—and ambiguous “post this” prompts should trigger a platform question first. During Launch you distribute launch notes and tag relevant topics; in Grow you sustain lifecycle conversations, search discussions, and follow builders in your niche. The reference stays lightweight because execution is native to the agent runtime; this file is the guardrails and API map indie operators need so automated posting stays on-brand and on-platform.910installs28Wechat BindingWeChat Binding lets you securely connect a user's WeChat account to your agent or app so it can push messages via send_to_wechat. You generate a QR code, the user scans and confirms in WeChat, and then exchange a bot token to establish the connection. Use this when setting up WeChat as a delivery channel, troubleshooting connection issues, or allowing users to connect and disconnect their accounts on demand—it's essential for any agent that needs to reach users on WeChat.767installs29Image EditDespite the image-edit slug, the documented capability is a cost-tracking helper for Starchild skill subprocesses that invoke sc-proxy via requests. Solo builders shipping paid media or API skills need every subprocess charge rolled into the same SSE cost_summary the main agent shows, not orphaned in logs. The module tags calls with SC-CALLER-ID, reads credit headers after each response, and appends rows under a default or overridden ledger path with file locking. It consumes STARCHILD_TOOL_CALLER_ID, STARCHILD_USER_TURN_ID, and optional ledger dir env vars set before bash dispatch. When those are missing, synthetic caller IDs still allow calls while audit rows remain for operators who later reconcile turns. Zero third-party dependencies make it easy to vendor beside image or other skills without bloating the agent runtime.2installs30Image CreateImage-create is a Starchild official skill for generating images through sc-proxy from a bash subprocess dispatched by your agent. It matters for solo builders who sell or demo AI products where every image request costs credits and users expect transparent per-turn accounting on the SSE cost_summary event. The bundled cost_tracking module is deliberately zero-dependency so you can copy the pattern into sibling skills: tag each paid request, record credits used, and let the orchestrator aggregate spend without coupling to starchild-clawd internals. Set STARCHILD_TOOL_CALLER_ID, STARCHILD_USER_TURN_ID, and optionally STARCHILD_COST_LEDGER_DIR before dispatch so ledger writes land under /data/.starchild/cost_ledger by default. Use during Build while you integrate generative media into workflows, landing experiments, or in-app assistants. Skip when you only need local Stable Diffusion scripts with no sc-proxy billing surface.1installs31Image TryonImage Tryon is a Starchild official agent skill pattern for virtual garment or product try-on backed by sc-proxy paid APIs. Solo builders shipping commerce or creative agents install it when they need try-on generation inside a bash subprocess without losing visibility into credits burned per user turn. A bundled stdlib cost-tracking helper ensures every requests-based sc-proxy call carries STARCHILD_TOOL_CALLER_ID and STARCHILD_USER_TURN_ID when the host agent sets them, then records ledger rows the agent merges into SSE cost_summary events. The skill fits builders already on the Starchild agent runtime who want generative media as a metered integration rather than a standalone script. Plan for network access to sc-proxy, review Security Audits on this Prism page before production, and treat ledger paths under /data/.starchild/cost_ledger as part of your ops story.1installs