or if you have any servers on Linux, must watch

your AI agent reverses correct answers when you push back. no evidence. just doubt. it folds. i built /meow to fix it. one word. four meanings. the agent picks which one from context. ► rechecking, when the agent made a claim worth testing ► continuing, when it stopped mid-task ► different angle, when it finished but missed ► picking, when it asked you to choose what it can decide works on Claude Code, Cursor, Codex, Aider, custom GPTs, raw API. MIT, no dependencies, 400 lines. 🛠️ repo: https://github.com/AgriciDaniel/meowmeow 📺 video: https://www.youtube.com/watch?v=Hz_SKQJ2KiE if you port it, tighten it, or catch a misclassification, drop it in the comments. i'll merge.

A Claude + Obsidian vault that builds and maintains a wiki as you work. v1.6 adds DragonScale: optional memory upgrades. 4 things you can turn on: 1. Log rollups (keep your log readable) 2. Stable page IDs 3. Duplicate-page detection (local, no API cost) 4. Auto-suggest what to research next, based on your vault graph Nothing changes if you don't opt in. https://github.com/AgriciDaniel/claude-obsidian/releases/tag/v1.6.0 Dragon Curve: https://en.wikipedia.org/wiki/Dragon_curve

This one's different. Six members of this community built skills and submitted them through the Pro Hub Challenge. Four brand new skills shipped. One enhancement. All of it is now in the official release. What's new: - Semantic Topic Clustering (by Lutfiya Miller, Challenge Winner) - Search Experience Optimization (by Florian Schmitz) - SEO Drift Monitor (by Dan Colta) - E-commerce SEO (by Matej Marjanovic) - International SEO Enhanced (by Chris Muller) 23 skills. 17 agents. 30 scripts. Free. Open source. Full walkthrough here: https://www.youtube.com/watch?v=1C6-ZadOHW8 Blog: https://claude-seo.md/blog/claude-seo-v190-community-release Pro Hub Challenge v2 is already live. Keyword: LEADS. $600 in Claude Credits. If you want your name in the next release, now's the time. ➡️ Join Pro

One command. /cybersecurity. 8 AI agents scan your entire codebase for vulnerabilities, missing auth checks, hardcoded secrets, supply chain risks, malware patterns and business logic flaws. I ran it on Claude Ads. Score: 62/100. Found SSRF vulnerabilities, unpinned GitHub Actions, missing CI gates. Fixed everything, pushed v1.5.1, score went to 90. Covers OWASP Top 10:2025, CWE Top 25, MITRE ATT&CK, 11 languages, 5 compliance frameworks. Framework-aware false positive suppression so it doesn't drown you in noise. Free. Zero config. Works immediately. GitHub: https://github.com/AgriciDaniel/claude-cybersecurity Video walkthrough If you're shipping code built with AI, run this before you publish.