ClawHub
PulseAugur coverage of ClawHub — every cluster mentioning ClawHub across labs, papers, and developer communities, ranked by signal.
9 day(s) with sentiment data
-
OpenClaw uses SSH tunnels for dashboard access and keyless remote command execution
This guide explains how to use OpenClaw, an open-source agent gateway, with SSH for secure access. It details two primary methods: accessing the OpenClaw dashboard via an SSH tunnel and enabling the OpenClaw agent to ex…
-
Security flaws found in combined AI agent skills, even when individual skills are safe
A new paper published on arXiv has identified significant security vulnerabilities in agent skills, even when individual skills are marked as safe. Researchers analyzed 13,598 skills from the ClawHub marketplace and dis…
-
AI agents vulnerable to email-based prompt injection attacks
A recent experiment demonstrated how AI agents can be manipulated through carefully crafted emails, leading to unauthorized command execution and data exfiltration without exploiting software vulnerabilities. This indir…
-
Graphify connects code knowledge graphs to OpenClaw, JetBrains, and Gemini CLI
Graphify has released plugins for OpenClaw, JetBrains AI Assistant, and Gemini CLI, enabling these tools to access a code knowledge graph. The OpenClaw and JetBrains integrations connect to Graphify Cloud via an API, al…
-
New Claude Code skills automate video and carousel creation for social media
Two new agent skills have been released for Claude Code, .codex, Cursor+, and OpenClaw, enabling users to automate the creation and posting of social media content. The 'Video to Shorts Skill' can transcribe long videos…
-
Orbital Desk launches API for 39,654 AI agent listings
A new API has been launched to catalog AI agents, providing a unified view of over 39,000 listings previously scattered across various directories. The API, developed by Orbital Desk, offers a free tier with 100 daily r…
-
AI agent network rejects its own promotion claim via independent model reviews
An agent network called ASIN-HHC, designed for third-party multi-agent coordination, subjected its own promotional claim to review by two independent AI models, Grok and Gemini. Both models rejected the claim, citing is…
-
ClawHub skill installs crypto miner, bypassing DNS security
A security analysis revealed a ClawHub skill, clawrtc, that installs a crypto miner on user machines. While the skill's disclosure omits that it sends hostnames and MAC addresses, it does request consent for data transm…
-
AI agent skills bypass security scans, violate policies
Researchers Taneja and Weber discovered a significant security vulnerability in AI agents, where skills that pass all security scans can still issue commands that violate an agent's established policies. Their analysis …
-
AI agent skills grant extensive capabilities, study finds
A new study analyzed 6,309 agent skills from two marketplaces, revealing that a significant portion grants extensive capabilities to AI agents. The research found that 70% of these skills possess at least one elevated c…
-
RollingGo offers free unlimited hotel MCP on ModelScope
RollingGo has launched a new hotel MCP (Model Communication Protocol) service available on ModelScope, offering developers free, unlimited access to a vast database of over 2 million global hotel properties. The service…
-
EPST offers developers email domain and API for AI agents
EPST is a new service that provides developers with their own email domain and a REST API to handle email-related tasks for their applications and AI agents. The service aims to reduce the time developers spend on repet…
-
AI agent skill registry aftermath analyzed: concentration and governance challenges
A recent paper examines the aftermath of the viral OpenClaw AI agent and its associated skill registry, which experienced rapid growth in early 2026. The study found that attention within the registry was highly concent…
-
Dromeas LLM Council Outperforms Claude Code Review on Complex PR
A comparison between two AI code review tools, Claude Code's ultrareview and Dromeas Code Review, found significant differences in their capabilities. When tested on a large, independently approved pull request from the…
-
Agent frameworks OpenClaw and Hermes clash over memory vs. channel control
Two competing open-source agent frameworks, OpenClaw and Hermes Agent, are vying for control of the agent runtime layer. OpenClaw, backed by major tech companies like Nvidia and Microsoft, focuses on broad distribution …
-
Claude Code skills audit finds 13% critical security flaws
A recent audit of Claude Code skills revealed significant security vulnerabilities, with over 13% containing critical issues and 36% exhibiting prompt-injection payloads. These malicious skills can exfiltrate sensitive …
-
ByteDance launches ArkClaw to monetize OpenClaw AI agent framework
ByteDance is developing a business strategy around its open-source AI agent framework, OpenClaw, by offering a cloud-based service called ArkClaw. This move aims to capitalize on the growing demand for AI agent tokens a…
-
Attackers poison Hugging Face with over 575 malicious skills
Attackers have poisoned Hugging Face and ClawHub with over 575 malicious skills originating from just 13 accounts. Separately, Google introduced the 'E' for Experience into its E-E-A-T guidelines in 2022, a concept prev…
-
Malicious actor embeds crypto mining in AI skill packages
A malicious actor has inserted cryptocurrency mining software into AI skill packages. This discovery highlights a security vulnerability within the distribution of AI tools and packages.
-
Study finds healthcare AI skills lack diagnostic focus, uneven coverage
A new study analyzes agent skills in healthcare, finding that current public skills primarily focus on patient-facing automation and monitoring rather than diagnostic or treatment tasks. The research, which examined 557…