CrawlForge MCP Server v4.8.0 Fixes Claude Skills That Never Auto-Activated

·1 views

CrawlForge released version 4.8.0 of its MCP Server on June 28, 2026, addressing a longstanding issue where Claude ignored skill files placed in the ~/.claude/skills/ directory. The update introduces seven auto-activating Agent Skills that use trigger-rich YAML descriptions, allowing Claude to independently match user prompts to the appropriate web-scraping tool without explicit instructions. The release also adds two new scrape formats — branding and screenshot — along with built-in scheduled change monitoring. Several security features that were previously advertised but silently non-functional, including SSRF enforcement and MCP elicitation confirmations, have now been properly implemented. The update is fully additive, leaving all 26 existing tools, their schemas, and credit costs unchanged for current users.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Drift Protocol Lost $285M to Six Months of Social Engineering, Not a Code Bug

On April 1, 2026, attackers linked to North Korea's Lazarus Group drained approximately $285 million from Solana-based DeFi platform Drift Protocol, making it the second-largest exploit in Solana's history. The attack involved no vulnerability in the smart contract code; instead, operatives spent six months building trust with team members who held admin keys before ultimately gaining access. A similar human-targeted attack hit KelpDAO two weeks later, resulting in a $292 million loss via a compromised LayerZero bridge developer. North Korea-linked actors were attributed to 76% of all crypto hack losses in early 2026, with private-key compromise now surpassing code exploits as the leading cause of theft. Security analysts stress that multisig controls, hardware key storage, time delays, and strict operational hygiene are now as critical as smart contract audits.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

MCP Server Tool Schemas Cost One Developer 42,000 Tokens Per Request

A software developer discovered that connecting a single GitHub MCP server added 42,000 tokens of schema overhead to every AI agent request, inflating token costs by 37% on the very first call. The Model Context Protocol requires each connected server to push its full tool schema into the model's context window at session start, meaning costs accumulate regardless of whether those tools are actually used. In a documented week-long test in June 2026 using Claude Opus 4.5, running four MCP servers with 47 tools consumed 80% of the context window on schema definitions alone, costing $8.92 per 1,000 requests versus $0.92 with no MCP. The developer applied four optimizations — including lazy schema loading triggered only when relevant and compressing tool descriptions from ~3,500 to ~800 characters — which cut token overhead from 80% down to 37%. The findings highlight a largely untracked cost factor for teams deploying AI agents with multiple MCP-connected tools in production environments.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Switches to Local AI Model After Claude Session Limit Corrupts Git Commit

A developer using a one-liner shell command to auto-generate Git commit messages via Claude AI ended up with the error message "You've hit your session limit" saved as an actual commit in their repository. The incident prompted them to explore running a local language model using Ollama, ultimately choosing the quantized qwen2.5-coder:1.5b model due to its low memory footprint of around 1.2 GB on an 8 GB RAM laptop. While the local setup worked without session limits, the developer discovered that Ollama restricts the default context window to 4,096 tokens based on available VRAM, causing the model to only process part of the diff and return inaccurate commit messages. Crucially, the model did not flag its own incomplete input — it responded confidently despite missing context, highlighting a key limitation of truncated inference. This led the developer to explore Modelfiles, a configuration layer in Ollama that allows tuning model parameters beyond prompt engineering alone.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Ex-Accenture BA Seeks Advice After Resignation, Job Search Struggles

A former Accenture Business Analyst has shared their career struggles on DEV Community after resigning from their first job due to challenges with spoken English and lack of project guidance. Assigned to a banking domain role requiring constant client communication, the individual found the demands overwhelming and eventually lost confidence. After four months of unemployment, interviews with TCS and Infosys also proved difficult due to the same communication barriers. The person has since been exploring Backend Development and Data Science but is uncertain which path to pursue. They are seeking community guidance on skill-building, job applications, internship opportunities, and rebuilding confidence in spoken English.

0 comments Read more at DEV Community