Tests AI agent systems for tool misuse, memory poisoning, hijacking and other agentic risks.
Haryy985/yunkun-console-preview
Gives a security agent ready-made playbooks for authorized penetration testing, from recon and API attacks to AI agent and MCP security.
- 31 skills
- 135 stars
The 5 most installed
Checks AI agent identity, role and permission boundaries for escapes, impersonation and credential leaks.
Tests API authentication and JWT handling for weak claims, header spoofing and rate-limit gaps.
Collects a website's API endpoints to map the attack surface.
The other 26 skills
Discovers API endpoints, schemas, versions and hidden documentation before testing.
Reviews REST, GraphQL, gateway, token and file APIs for security weaknesses in one pass.
Tests payment, order, coupon and approval flows for business logic flaws and state abuse.
Checks Java, PHP and Python apps for unsafe deserialization that could lead to code execution.
Tests file upload features for bypass tricks that allow malicious files through.
Tests GraphQL endpoints for introspection leaks, injection, authorization bypass and query abuse.
Tests MCP servers and tools for poisoned descriptions, hidden instructions and token theft.
Tests OAuth, OIDC and SSO login flows for redirect hijack, CSRF and token leaks.
Automates browser interactions and runs Playwright tests on web pages.
Tests AI systems that read external data for indirect prompt injection and data leaks.
Tests web apps for race conditions, concurrent abuse and rate-limit bypass.
Maps assets, fingerprints technology and builds a structured testing plan for a target.
Tests for HTTP request smuggling and desync between proxies, CDNs and origin servers.
Manages ScopeSentry scanning projects, tasks, templates and assets through its MCP.
Plans systematic security testing with a layered thinking model and OWASP/WSTG mapping.
Routes an authorized security workflow across collection, profiling, validation and evidence replay.
Collects and normalizes assets, routes, APIs and technology into an evidence graph before testing.
Tests inputs that reach SQL queries for injection, including blind and out-of-band paths.
Runs one bounded discovery hypothesis at a time and returns results as evidence.
Turns authorized security findings into well-evidenced reports with severity ratings and reproduction steps.
Tests URL-fetching, webhook and PDF features for server-side request forgery and internal network access.
Maps a target's technologies, trust boundaries and attack surface to prioritize what to test first.
Checks WebSocket handshakes and real-time channels for cross-site hijacking and other common flaws.
Finds and proves reflected, stored and DOM cross-site scripting, including WAF and CSP bypasses.
Plans authorized security validation tasks and produces HTTP packet evidence and reviewed reports in Yunkun.