Agent entry
Let an agent find, trust, and install the right skill.
This page is the low-noise map for agent browsers. Start with a task, resolve the best skill, inspect trust signals, then fetch an install handoff.
Recommended flow
How an agent should use this registry
01
Read /llms.txt
02
Choose a task
03
Call Resolve API
04
Fetch install handoff
Task layer
High-intent task routes
Task pages are designed for agent browsers and search traffic. They include best first install, alternatives, and decision notes.
web-scraping
Scrape pricing pages
Extract pricing data from public competitor pages and turn it into clean structured output.
web-scraping
Crawl docs
Turn documentation pages into clean markdown or records that an agent can search and reuse.
coding-agents
Review PRs
Help a coding agent inspect diffs, summarize risk, and suggest targeted review comments.
testing-qa
Fix CI
Debug failing checks, inspect logs, patch likely causes, and verify the fix.
document-processing
PDF to markdown
Extract readable text, tables, and metadata from PDFs for agent workflows.
rag-knowledge
Build RAG
Ingest docs, chunk content, retrieve relevant passages, and cite sources in agent answers.
data-analysis
Analyze CSV
Load tabular files, profile columns, find patterns, and produce concise analysis.
browser-automation
Browser workflow
Navigate web apps, fill forms, take screenshots, and verify state through browser automation.
Agent endpoints
Stable surfaces for tools
/api/agent/resolve
Task to selected skill, alternatives, policy decision, safety profile, and install plan.
/api/agent/tasks
Structured task definitions with Resolve API URLs.
/api/agent/skills/crawl4ai
Agent-readable skill profile with trust, audit, install targets, and URLs.
/api/skills/crawl4ai/install?format=text
Plain text command, agent prompt, and safety checklist.
/.well-known/agent-manifest.json
Machine-readable site contract for agent discovery.
/openapi.json
API schema for tools and agent runtimes.
Auto-discovery
GitHub collection is already automated
Production cron calls the indexer hourly. It imports high-star skill-like GitHub projects, excludes MCP-only projects, records runs, and refreshes stars daily.
Top registry records
High-signal skills for agent routing
Crawl4AI
Open-source LLM-friendly web crawler and scraper
66,195 stars
AutoGPT
Build and run autonomous AI agents for open-ended tasks
184,887 stars
MarkItDown
Convert documents into Markdown for agent-readable context
146,460 stars
LangChain
Agent engineering primitives for tools, memory, and RAG
138,996 stars