ski

Anthropic announced that it will donate the Agent Skills standard to the Agentic AI Foundation (AAIF) under the Linux Foundation. This format was designed by Anthropic while developing Claude Code, set to launch in October 2025 and open as a cross-platform standard in December. Currently, over 70 agent platforms, including ChatGPT, Cursor, and GitHub Copilot, have integrated it.Anthropic describes Skills as reusable work manuals prepared for AI agents, where developers can write task steps in SKILL.md and attach scripts and reference materials for different agents to read as needed. MCP is responsible for connecting external tools and data, while Skills explain how agents complete specific tasks. What is being donated includes the standard name, logo, GitHub organization, and official website domain.AAIF was jointly initiated by Anthropic, OpenAI, and Block in 2025 and has previously received MCP. In the future, engineers from Anthropic, OpenAI, and Google will jointly maintain the Skills standard in their personal capacities, no longer solely led by Anthropic. Vercel statistics show that the skills.sh directory has recorded approximately 1 million skills, with nearly 280 million installations. Snyk has confirmed 76 malicious skills in ClawHub's skill samples, and Anthropic has proposed establishing version control, formal proposal processes, and security guidelines.

According to "Du Jia," the Meituan AI team has launched the AI Agent skill platform DeepSkill, which currently includes over 370,000 validated Skill tools that have been tested through real tasks, nearly doubling from 130,000 at the beginning of September.The platform does not use a traditional directory search model; users can directly describe their task requirements, and the platform automatically breaks down and matches candidate Skills from the library. The platform will scan and analyze all content of the Skills, outputting an analysis report that includes capability boundaries, tool invocation, potential risks, and data security reviews. It also features trend rankings on GitHub, authoritative source rankings, and market hot search rankings. Currently, major companies such as Tencent, ByteDance, and Alibaba have made arrangements in the Skill platform field.

Despite the S&P 500 Index closing at a historic high this week, the rally was primarily driven by companies benefiting from the artificial intelligence boom, overshadowing signs of weakness in the U.S. economy and financial markets. Companies like Meta, Microsoft, and Nvidia performed strongly, while stocks in the consumer, travel, transportation, real estate, and financial sectors declined.When the S&P 500 Index reached its historic high, only 30% of its constituent stocks were trading above the 50-day moving average.

Famous trader Killa stated that he does not believe Bitcoin will hit new lows again and is currently maintaining a long position. However, after the U.S. midterm elections, the market may still experience some de-risking behavior, so it is not the right time to establish hedge short positions in advance.Historically, Bitcoin's performance has been weak after every U.S. midterm election cycle. However, if Bitcoin continues to rise before the elections, traders can wait until the midterm elections are closer to decide whether to take action based on the price trends at that time.He believes that a more reasonable strategy right now is to continue holding long positions and observe market reactions before and after the midterm elections, and then consider whether to establish hedge short positions, rather than shorting in advance solely based on historical cycles.

The criminal trial of Aiden Pleterski, who calls himself the "King of Crypto," regarding allegations of cryptocurrency fraud will begin on October 5 in the Ontario Superior Court of Canada. He is accused of raising funds from investors between 2021 and 2022 by making false promises of "smart investments yielding high weekly profits," with the fraud amounting to approximately $30 million. Canadian authorities have filed charges of fraud and money laundering against him in May 2024.According to CBC News, Pleterski failed to obtain a postponement to hire a lawyer and is expected to represent himself in court. The trial is expected to last about four weeks, and if convicted on the fraud charge alone, he could face up to 14 years in prison.In December 2022, several men (some of whom are reportedly victims of the scam) allegedly kidnapped and tortured Pleterski for three days, threatening his and his family's safety, and forcing him to record a video detailing his role in the alleged fraud. Pleterski is one of many high-profile figures in the cryptocurrency industry facing criminal charges during the market downturn in 2022, with executives from the now-bankrupt FTX and Celsius also facing charges in the United States thereafter.

a16z Charts compiled recent data indicating that since January 2025, tech recruitment has shown a clear preference for years of experience, while the requirements for the number of skills have decreased, with a change of about 5% to 10%.The charts show that companies are lowering the weight of skill lists when screening positions, and are instead placing more emphasis on candidates' accumulated experience. This change is related to the skill gaps brought about by AI, which is rewriting the safety and employment landscape.

BSC and 0G ecosystem AI economic infrastructure project NeoSoul announced that its general intelligent trading workstation NeoTrade has launched the trading Skill Builder. Users only need to describe their trading ideas or trading styles in a conversation, and the system can generate the corresponding trading Skill, automatically completing the configuration and verification of strategy parameters such as time period, market focus, position, and order method.In the past, traders needed to turn their trading ideas into strategies that Agents could execute sustainably, which often required writing trading logic and setting parameters item by item. The Skill Builder centralizes the process into a conversation window, allowing users to describe their core trading ideas in natural language, and then gradually refine the strategy through continuous dialogue. During the conversation, NeoTrade will configure specific conditions based on user requirements and write each modification in real-time into the Skill. Once the strategy is complete, the system will automatically perform parameter verification, and the generated Skill can be saved and reused.Kaelan, co-founder of NeoSoul, stated that turning trading judgments into rules that Agents can execute has always been one of the most time-consuming steps in creating trading Agents. The Skill Builder incorporates this step into a single conversation, where users first clarify how they want to trade, and the system is responsible for completing the structuring, parameter configuration, and verification.

Numa Lunah, co-founder of the crypto project Refi Hub, stated that he was hacked after installing a transcription application using a download link provided in the Claude chat window. The link pointed to a counterfeit website and bundled malware that attempted to steal all information from his device upon execution. Numa Lunah mentioned that he cleared and reinstalled the affected laptop and found no sensitive information leaked.Subsequently, he discovered a contaminated Claude Code skill file SKILL.md in his backup, which disguised itself as a style guide written by him and contained instructions to re-download the malware and steal credentials every time it was loaded. Microsoft Defender Experts had warned that attackers had shifted from search engine optimization poisoning to large language model response poisoning, with related tactics including recommending download links controlled by attackers, AI brand impersonation installers, and contaminated code libraries and proxy skills. Crypto industry practitioners may hold irrevocable credentials such as mnemonic phrases, private key files, hot wallet JSON, exchange API keys with withdrawal permissions, and deployer keys.

At a closed-door sharing session held in Hong Kong, Binance founder CZ responded to questions about entrepreneurship among university students, emphasizing that there is no need to rush into starting a business. He suggested that young people should first work at a good large company for a few years, stating, "Two years should be the minimum," with 2 to 5 years being ideal. He cited his own four years at Bloomberg as an example, noting that the complete management system of a large company is something that university students cannot experience immediately upon graduation, which will be very helpful for future entrepreneurship.CZ believes that young people must focus on cultivating "a specialized skill." He stated that this is a world of global competition, "If you broadly say programming, there are millions of programmers worldwide, and it is impossible to be twice as good as everyone else. However, if you can become the best 1% or even 0.1% in the world, the returns you receive may be several times, dozens of times, or even hundreds of times more than others. The treatment of the first place and the second place will be very different."

According to CoinDesk, citing informed sources, the cryptocurrency custody company Copper, which was once valued at up to $2.5 billion, has attracted the interest of potential buyers, but the offers are only $200 million, far below the $500 million offer made by Cantor Fitzgerald a few months ago.Informed sources revealed that the company is facing challenges related to preferred stock after completing over $300 million in financing at a high valuation.

Recently, Peter McCrory, the head of economics at Anthropic, explained the AI unemployment paradox. He pointed out that AI capabilities have significantly improved, raising reasonable questions about why there has not yet been a noticeable impact on unemployment rates. The U.S. Business Trends and Outlook Survey shows that about one-fifth of U.S. businesses use AI to some extent, and productivity growth in the U.S. has strengthened in recent years, some of which may be related to AI.McCrory stated that the unemployment rate in the labor market is still close to the full employment level identified by the Federal Reserve, and even deep data analysis makes it difficult to draw clear conclusions about the impact of AI. The basic judgment is that so far, AI has shown characteristics of skill-biased technology, automating certain work processes while complementing and amplifying the professional skills that rely on human participation, thereby creating maximum value in human-machine collaboration.

The Web3 asset data platform RootData has now fully opened its personnel data, adding data on current and former team members in its project query function published on ClawHub Skill. It has newly added dynamic data on positions in the crypto field (covering the latest 200 hiring/termination updates) and supports queries for information on over 20,000 individuals in the crypto field, including introductions, links, and work details. All data is available for free access by AI agents.RootData has long tracked tens of thousands of LinkedIn and Twitter data, and based on user-submitted data, it maintains the most comprehensive and timely personnel database in the industry to meet users' more detailed information needs and reduce information barriers in the industry.RootData stated that through this comprehensive opening, it hopes to help more users quickly find key talents, gain insights into industry talent flow trends, and provide efficient data support for Web3 project recruitment, investment decisions, and industry research.

The National Internet Emergency Center (CNCERT) stated that some skill packages for large models are being disseminated under names like "large model jailbreak" and "mining for profit," which are used to bypass model security restrictions or induce users to perform mining operations.The report specifically names Skill "godmode," which achieves jailbreaking through methods such as replacing system prompts and input obfuscation, potentially leading to the generation of illegal content, triggering service bans for large models, and risks of privacy leakage; Skill "Bonero-Miner" guides users to download mining programs to mine the anonymous Bonero coin, posing risks of being involved in money laundering and economic losses due to high device load.CNCERT advises individuals and enterprises to obtain Skills only from official channels, strictly control permissions and access whitelists, and regularly clean up suspicious components.

Lighter may benefit from the recent "fireside chat" between founder Vlad Novakovski and Vitalik discussing high-performance applications on Ethereum and the ZK Rollup scaling roadmap, as well as smart money seeking undervalued targets in the same sector amid multiple favorable conditions. Its platform token Lit has strengthened against the overall weakness in the crypto market. According to HTX data, it is currently priced at $1.771, with a 24-hour increase of 21.87%. The current market focus is on Lighter's high-performance application scenarios in the Ethereum ecosystem, its scalability roadmap, and the future development direction of decentralized finance (DeFi).In addition, according to DefiLlama data, Lighter's 24-hour trading volume reached $1.289 billion, with a holding volume exceeding $760 million.

Elon Musk's AI company xAI announced the simultaneous launch of the "Skills" feature on the web, iOS, and Android platforms. This feature gives the AI assistant Grok the ability to have persistent memory across conversations, marking Grok's evolution from a traditional one-time Q&A chatbot to a configurable, programmable automation workspace.According to the introduction, with the "Skills" feature, users only need to teach Grok the operation method for a specific task once, and Grok will permanently remember relevant personal preferences, formatting rules, or specific workflow steps in all subsequent independent conversations. This means that users no longer need to repeatedly input cumbersome pre-prompts or background settings when starting a new conversation.

Chainalysis posted on the X platform that before the theft of THORChain, wallets suspected to be associated with the attacker had been transferring funds through Monero, Hyperliquid, and THORChain for several weeks. The attacker-associated wallets had already deposited into Hyperliquid positions via the Hyperliquid and Monero privacy bridge as early as the end of April. The funds were then exchanged for USDC and transferred to Arbitrum, and later bridged to Ethereum, with some ETH subsequently transferred to THORChain to become staked RUNE for newly added nodes, which are believed to be the source of the attack.Afterward, the attacker bridged some RUNE back to Ethereum and split it into four pathways, one of which went directly to the attacker. After being transferred through intermediate wallets, 8 ETH was sent to the final wallet receiving the stolen funds 43 minutes before the attack. The funds from the other three pathways flowed in the opposite direction. These wallets bridged ETH back to Arbitrum, deposited it into Hyperliquid, and transferred it into Monero through the same privacy bridge, with the last transaction occurring less than 5 hours before the attack began.As of Friday afternoon, the stolen funds have not yet been used, but the attacker has demonstrated their skilled cross-chain money laundering capabilities, and the Hyperliquid to Monero path may become the next move.

According to Slow Fog CISO @im23pds's post on the X platform: attackers used 13 accounts to implant 575 malicious Skills into Hugging Face and ClawHub.