Browse and search
Start from the public catalog, filter by task fit, resource type, verification status, capability tags, scan signals, and community feedback.
Platform overview
How Agentique works
Agentique is a free sharing hub for AI-agent resources. It helps users discover, inspect, upload, scan, download, and discuss reusable agents, skills, workflows, bundles, and examples while keeping safety status and platform boundaries visible.The user path is inspectable end to end.
Each step is designed around reuse decisions: what the resource is, who published it, what it can touch, what the scanner saw, and what the user should confirm before adoption.
Inspect before reuse
Resource pages expose manifests, package references, hashes, compatibility notes, distribution mode, publisher signals, and safety readback before download.
Upload with checks
Publishers submit metadata, source links, package details, capability declarations, and examples through a guarded wizard instead of a raw file drop.
Scan and review
Uploads are treated as untrusted input. Static analyzers, unsafe artifact checks, prompt-risk checks, quarantine states, and review gates run before public reuse.
Download with context
Download handoff keeps version, checksum, scan evidence, source status, and non-execution disclosure attached so adoption is deliberate.
Interact and improve
Users can react, review, comment, collect, share, and follow updates so resource quality improves through visible community feedback.
Resource types stay explicit.
Agentique treats agent resources as structured records rather than anonymous files, so users can compare capabilities, compatibility, source, and safety posture before reuse.
- Agent profiles with manifests and task-fit metadata
- Skill packages with strict validation and capability declarations
- Workflow templates with sanitizer readback
- Bundles that group compatible resources without hiding individual risk
- MCP-ready and protocol-shaped metadata while public protocol surfaces remain gated
- Examples, articles, and discussion records that explain practical use
Safety scanning is a platform responsibility.
Agentique acts as a supervisory layer for shared resources. The scanner path assumes uploads may be malicious or misleading, then records readable evidence for users and reviewers.
Fail closed before public reuse
Suspicious packages can be quarantined for review. High-risk capabilities, unclear source claims, unsafe scripts, and workflow changes are surfaced as readback instead of hidden behind a generic trust badge.
- No-execution intake for untrusted uploads
- Static analysis for risky scripts, unsafe artifacts, and dependency signals
- Prompt-injection, tool-poisoning, credential-harvesting, and permission-risk readback
- Hash, source, package, and manifest evidence persisted with the resource
- Quarantine and human-review checkpoints for suspicious submissions
- Agent-mediated consent language for high-risk actions and external handoffs
Trust is built from multiple signals.
A resource can be useful and still risky. Agentique separates identity, package, workflow, scanner, and community evidence so users can make a narrower decision.
Publisher identity
Creator pages, verification lifecycle, namespace claims, and social/source signals make provenance easier to compare.
Package clarity
Compatibility, distribution mode, hashes, package references, and bundle membership stay visible instead of being collapsed into a single download button.
Workflow readback
Workflow resources show sanitized steps, capability boundaries, and reviewable changes before users adopt or adapt them.
Safety status
Scan state, review state, moderation signals, and unsafe-artifact warnings are part of the resource record.
Boundaries are part of the product.
Clear limits keep the shared catalog useful without pretending that every deployment, protocol, or runtime concern is already open to the public.
Free sharing catalog
Agentique is positioned as a free community sharing platform. Paid marketplace, escrow, bidding, and revenue-share mechanics are not part of the current product path.
Not hosted execution by default
Resources are inspected, described, downloaded, and discussed. Running third-party agents as hosted workloads remains gated behind explicit future safety and deployment controls.
Protocol surfaces stay controlled
Public and protocol APIs are separated from first-party app surfaces. External automation routes remain feature-gated until production readiness is proven.
User consent before risky handoff
Capability disclosure, download readback, and high-risk warnings are designed so agents and users can pause before adopting external resources.
Built for shared learning, not one-off downloads.
Profiles, personal libraries, notifications, collections, reviews, reactions, issue context, and creator pages keep resource adoption connected to ongoing community feedback.