Apify Agent Skills
Give AI coding agents expert hands on Apify's 30,000+ Actors for web scraping and automation.
- Source repo
- apify/agent-skills
- Stars
- ★ 2.4k
- Last updated
- 2d ago
- Primary language
- Python
- FA score
- 40/100 · Major gaps
At a glance
- How it runs
- Works with
- Portable with changesCodex · Claude Code
- Cost
- Free tier plus a paid hosted plan
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A sales or marketing operator asks the agent to find Brooklyn restaurants rated under 4 stars on Google Maps, scrape their reviews and contact pages, and export a ranked CSV for the CRM.
- Not a fit if
- Teams that need fully offline operation with no external data calls
- Users unwilling to create an Apify account or pay per-Actor usage
- Developers who only need code editing, not web data collection
- Source review
- 40/100 · Major gaps 1 safety controls not found
What does this agent do, and when should you use it?
Apify Agent Skills is an official Apify collection of Markdown skill packs that teach AI coding agents how to use the Apify platform for web scraping and automation. The repository ships four core skills: apify-ultimate-scraper (130+ curated Actors covering Instagram, TikTok, YouTube, Google Maps, Amazon and more, with automatic fallback to the full Apify Store of 30,000+ Actors), apify-actor-development (create, debug and deploy Actors in JavaScript, TypeScript or Python and generate their input/output schemas), apify-actorization (wrap existing scripts, libraries or CLI tools as Actors), and apify-integration-development (build official Apify integrations, AI framework packages or direct clients). A commands/ pack adds slash commands such as /create-actor for guided scaffolding. Installation works through npx skills add, the Claude Code plugin marketplace, or by referencing the Markdown files directly (agents/AGENTS.md, skills/<name>/SKILL.md). Running it requires an Apify account, the Apify CLI or APIFY_TOKEN, and Node.js 20.6+; the actual scraping executes on Apify's cloud Actors and is billed per usage.
Each skill is a Markdown instruction bundle (SKILL.md). Once loaded by Claude Code, Cursor, Windsurf, Codex or Gemini CLI, the agent follows the documented flow: pick an Actor, shape its input, launch a run through the Apify CLI (apify login or APIFY_TOKEN), poll run status, and format results as CSV or other structured output. apify-ultimate-scraper matches the target platform against 130+ curated Actors and falls back to searching the full Apify Store when nothing fits. apify-actor-development generates or updates actor.json, input_schema.json, dataset_schema.json, output_schema.json and key_value_store_schema.json, and bundles references for README, logging and standby mode. apify-actorization converts existing code using the JS/TS SDK, the Python async context manager or a generic CLI wrapper. apify-integration-development supplies per-category capability matrices and definition-of-done checklists for workflow-automation apps, AI agent harness plugins, AI framework packages and direct application clients.
- A sales or marketing operator asks the agent to find Brooklyn restaurants rated under 4 stars on Google Maps, scrape their reviews and contact pages, and export a ranked CSV for the CRM.
- A competitive-intelligence analyst pulls pricing pages, G2 and Trustpilot reviews, job postings and social posts for three competitors and summarizes positioning gaps.
- A Python or TypeScript developer wraps an existing scraper script or CLI tool as an Apify Actor using apify-actorization.
- A SaaS team building an official Apify integration follows apify-integration-development's checklists for a Zapier/n8n-style or LangChain/LlamaIndex-style package.
- A brand team collects 30 days of brand mentions across Instagram, LinkedIn, X and YouTube and runs sentiment analysis for top complaint and praise themes.
- An engineer already using Claude Code or Cursor installs the plugin once and lets the everyday coding assistant call Apify Actors on demand.
How do you install or deploy this agent?
Prerequisites: an Apify account (free tier available), the Apify CLI and Node.js 20.6+. Install the CLI and authenticate:
npm install -g apify-cli
apify loginIn headless or CI environments without a browser, generate a token in Apify Console → Integrations and set it as an environment variable instead:
APIFY_TOKEN=your_token_hereOption 1, install a single skill with the skills CLI:
npx skills add https://github.com/apify/agent-skills --skill apify-ultimate-scraperOption 2, install via the Claude Code plugin marketplace:
/plugin marketplace add https://github.com/apify/agent-skills
/plugin install apify-ultimate-scraper@apify-agent-skills
/plugin install apify-actor-development@apify-agent-skills
/plugin install apify-actorization@apify-agent-skills
/plugin install apify-integration-development@apify-agent-skillsCursor and Windsurf reuse the Claude Code plugin format. For Codex CLI and Gemini CLI, clone the repository and point the agent at agents/AGENTS.md or gemini-extension.json:
git clone https://github.com/apify/agent-skillsHow do you use this agent?
After installation, describe the outcome in plain language; the agent selects Actors, shapes input, runs them and formats results. For example:
Scrape the top 50 results for "AI coding tools" from Google Maps and save them to a CSV.Without the plugin route, reference the skill files in your agent's context: agents/AGENTS.md is a one-page index of every skill, and skills/<skill-name>/SKILL.md holds the full instructions. The commands pack adds guided scaffolding, e.g. type:
/create-actorNote that apify-generate-output-schema is deprecated; its functionality now lives in apify-actor-development.
What are this agent's strengths and limitations?
- apify-ultimate-scraper ships 130+ curated Actors across major social, search, maps and e-commerce platforms, with automatic fallback to the full 30,000+ Actor Apify Store.
- Skills are agent-agnostic SKILL.md Markdown, so the same content works in Claude Code, Cursor, Windsurf, Codex and Gemini CLI.
- Beyond scraping, the repo covers Actor development, code actorization and official integration building, spanning data acquisition to delivery.
- apify-actor-development generates or updates input_schema.json, dataset_schema.json and related schemas, cutting manual configuration.
- All scraping runs on Apify cloud Actors behind an Apify account; nothing works fully offline, and Actors bill separately per result or per event.
- Requires Node.js 20.6+ plus the Apify CLI or APIFY_TOKEN, and headless environments must manage the token manually.
- The README marks apify-generate-output-schema as deprecated, so existing users must migrate to apify-actor-development.
- Execution quality depends on how faithfully the host agent follows Markdown instructions; the repo provides no automated test evidence.
How does this agent compare with similar options?
The README points to apify/awesome-skills, a community-maintained collection of domain-specific skills (lead generation, brand monitoring, competitor intel). It complements this official general-purpose collection: teams with a narrow vertical need may prefer it first.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Apify Agent Skills This agent | 40 · Major gaps | Agent plugin / skillFreemium | ★ 2.4k | 2d ago | Python | Codex · Claude Code |
| Apify MCP Server | 77 · Some gaps | Hosted serviceFree + model costs | ★ 12k | today | TypeScript | ChatGPT · Claude Code · Claude.ai |
| Firecrawl: Web Data API | 41 · Major gaps | Library / SDKFreemium | ★ 194k | today | TypeScript | ChatGPT · Claude Code |
| Maxun | 48 · Major gaps | Hosted serviceFreemium | ★ 18k | today | TypeScript | — |
How does FollowAgents rate this agent?
Why each dimension lost points
README indicates the skills drive Apify Actors to scrape external sites and require APIFY_TOKEN, implying external side effects and credential use; however the repository evidence contains only README and one CI workflow, with no SKILL.md bodies, least-privilege guidance, pre-run user confirmation, data-flow disclosure, sensitive-data handling, or rollback mechanism. Dependency security is limited to CI using actions/checkout@v4 and setup-uv@v4 without pinned hashes or audit. Source attribution rests on badges and brand claims with unverified publisher identity. Scores of 1 reflect thin/asserted coverage; rollback is 0 for absent evidence.
README is internally consistent (skill count, install commands, prerequisites align) and CI enforces AGENTS.md/marketplace.json sync, showing some consistency control; but dependency availability is only a loose Node.js 20.6+ and apify-cli requirement without pinned versions, and failure messages/error handling are not described in visible files.
README targets multiple hosts (Claude Code, Cursor, Windsurf, Codex, Gemini CLI) and lists many scenarios, giving good audience/scenario coverage; capability boundaries are marketing-style with no guidance on when not to use or scraping compliance limits; trigger precision depends on SKILL.md which is not in evidence; environment fit is supported by install and prerequisite notes, scoring 2.
README is well structured with anchors, install steps, examples, and a license badge, so information architecture and install notes score well; naming stability is shown by deprecating apify-generate-output-schema with a migration note; but known limitations are largely absent, versioning/changelog has no substance, and maintenance responsibility is only self-asserted by an unverified publisher, hence deductions.
README shows output shapes (CSV, structured results) and use cases but lacks actual skill files, output samples, and cost quantification; marginal value over directly using Apify Store/MCP cannot be judged from evidence; cost is only described as pay-per-result/event with free credits and no estimates, so all score 1.
Most capability claims (130+ curated Actors, 30,000+ Actors, 5 skills) have no supporting files and trace only to brand blog and external links; cross-source corroboration is limited to CI validating AGENTS.md, which does not verify skill content; separation of fact from inference is weak as marketing claims mix with verifiable facts.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Repository evidence contains only README and one CI workflow; SKILL.md bodies are absent, so actual skill behavior, permission scope, and triggers cannot be verified.
- Skills use APIFY_TOKEN and drive external scraping, posing credential and external-effect risk, yet no least-privilege, user-confirmation, or data-flow documentation is present.
- Publisher identity is unverified; maintenance responsibility and update path are self-asserted in README and should not be used to infer reliability.
- Dependencies are not pinned (Node.js 20.6+, apify-cli unpinned) and CI actions lack hash pinning, introducing supply-chain uncertainty.
- Known limitations, changelog, and rollback are missing; assess scraping compliance and cost independently before production use.