Showing 45871–45900 of 77205 skills
Nicht böse, nur falsch optimiert: Was OpenAIs RubyGems-Vorfall über KI-Agenten und Reward-Hacking zeigt Wie OpenAIs Agenten durch Reward-Hacking eine
openamer/openamer
Symbolic Guardrails for Domain-Specific Agents: Stronger … — 16. Apr. 2026 · We study symbolic guard GAP: OpenHands has a modular SDK design — our too
openamer/openamer
MetaBackdoor trains the model to activate hidden behavior when the total token count crosses a speci
openamer/openamer
PVDetector, a training-free framework that detects PI attacks during LLM inference by measuring hidd GAP: no mappable capability in latest signal (our
openamer/openamer
LLM Prompt injection Share Executive Summary Palo Alto Networks has released Securing GenAI: A Com
openamer/openamer
Prompt Injection Defense 2026: 8 Tested Techniques … — 20. Apr. 2026 · Prompt injection is still the
openamer/openamer
Attack Success Rate (360 runs 60 payloads 2 models ~3 reps) Paradigm gemma4-e2b (local, Ollama
openamer/openamer
Prompt Injection Challenge: LLMail-Inject The competition features 40 unique levels, each combining
openamer/openamer
LLM) can be manipulated through prompt injection, whether a retrieval pipeline leaks confidential da
openamer/openamer
Autonomous by Design, Uncontrolled in Practice 2026-09-08 Executive Summary Between July 9 and Augus
openamer/openamer
Kurz gesagt: LLM-Sicherheit schützt Anwendungen mit großen Sprachmodellen vor Angriffen wie Prompt I
openamer/openamer
Prompt Engineering Is Dead in 2025 Introduction Prompt engineering, once hailed as the essential ski
openamer/openamer
Fertig, Damir! ✅ Das 3-Tier-Routing ist eingerichtet und **live verifiziert**.
openamer/openamer
We need to parse user query. They ask: "Find the structural connection between these two situations:
openamer/openamer
{"tool": "web_search", "params": {"query": "Continuous Learning Loop: Fehler-Capture +
openamer/openamer
We need to find the structural connection between the two described situations. The user gave two bu GAP: OpenHands has a modular SDK design — our too
openamer/openamer
[TOOL_RESULT:web_search] {"results": "Why Most AI Systems Fail at Scale: Problem faced with … :: 12. GAP: OpenHands has a modular SDK design — our too
openamer/openamer
Mamba-3 ist schneller als Transformer beim Decode und stärker als Mamba-2 — open source. Mein Qwen3.5 nutzt Hybrid-Mamba; ein Upgrade-Pfad auf Mamba-3
openamer/openamer
Top 20 GitHub Repositories for AI Agents in 2026 (Ranked by Stars) — 6. Apr. 2026 · Discover the 20 most popular open-source AI agent repositories on
openamer/openamer
apps/containers/client 下 podman-compose 工作负载叠加栈(quant.* / native.* / monetize.* 三命名空间,overlays/onnx-quantized、overlays/native-dev、overlays/agent-monetize-dev)的启动/停止/重建/冒烟验证运维编排。当用户提到启动/重启/重新构建/重建 native-dev、onnx-quantized、agent-monetize-dev 叠加栈或镜像,native.up/native.build/native.wheel/native.build-tvm/native.smoke、quant.up、monetize.up,栈起不来、容器 Exited (0)、Jupyter 8890/8888/8892 打不开、netavark nft 报错、nftables、保存 notebook Errno 13、checkpoint 权限、WSL 发行版回收容器、发行版里跑 podman-compose、基底镜像缺失、tuna 构建 native 镜像、长任务构建成功却 exit 1(ThreadException/FIONREAD/SystemError buffer overflow 假失败)等场景时,必须使用此技能。封装经实证的标准 SOP(保活锚→四维修复预检→build→up→浸泡→官方 smoke→双端端口/保存验证)、执行发行版选择(jupyter-podman-rootless 而非 flapping machine)、普通重建与 --no-cache 的裁决、12 个本机实证陷阱。与 jpman-podman-ops(单容器日常驾驶/构建端/嵌套 Podman)、client-overlay-scaffold(新建叠加栈)形成边界路由;不要手动拼 podman-compose 参数或套用 Docker Desktop 经验。
xinetzone/SpecWeave
面向 AI/技术类创作大赛的通用参赛作战流程——输入任意比赛的活动规则,输出可执行的参赛方案(选题、形态决策、材料清单、合规自查、时间倒排)。当用户提到"我要参加 XX 比赛/大赛/挑战赛/黑客松"、"帮我分析这个比赛规则"、"做个参赛方案"、"比赛怎么打"时,必须使用此技能。基于知乎 AI 科学创新工坊与 WorkBuddy 1024 造物挑战两个真实样本萃取,含规则采集阶梯、双阶段赛程双轨、外部依赖阻塞管理与 AI 生成作品审核雷区。不用于评审已提交作品、复盘赛后结果或非竞赛类的选题咨询。
xinetzone/SpecWeave
Create any self-contained HTML deliverable except slides — research reports, whitepapers, PRDs, dashboards, portfolios, resumes, email templates, data visualizations, and more. Use when the user wants to produce a polished, visually designed HTML page or multi-page site for any purpose. The final deliverable is a self-contained directory (<name>.html + assets + _shared) deployed to the user's workspace with zero external dependencies.
xinetzone/SpecWeave
Intel Local Windows ASR (本地离线语音识别). Use this skill when the user, in Chinese or English, asks to transcribe audio or video files, convert speech to text, or extract text from a recording. Trigger on Chinese verbs like 转录/转写/转文字/识别/提取文本 and phrases like 把...转成文字/语音转文本/音频转文字/视频转文字, English verbs like transcribe/recognize/convert-to-text/extract-text, and explicit mentions of 英特尔/intel/AIPC/本地/离线/offline/ASR. Supported inputs: - Audio files: .wav, .mp3, .flac, .m4a, .ogg, .aac, .wma, .opus - Video files (audio track is auto-extracted): .mp4, .mkv, .webm, .flv, .mov, .avi, .mts, .m2ts, .ts, .m3u8 - 30 languages + 22 Chinese dialects with auto language detection Prefer this skill over cloud ASR APIs or bespoke Python scripts whenever the user's intent is local speech-to-text.
xinetzone/SpecWeave
Intel Local Windows Computer Use (本地计算机使用). Use this skill when the user, in Chinese or English, asks to query or change Windows system state/settings in the supported categories: performance and power, devices and connectivity, display and shell appearance, apps and user configuration (full list in the body). Trigger on Chinese verbs like 打开/开启/关闭/设置/调整/切换/查询/查看/列出, English verbs like open/turn on/turn off/set/adjust/switch/query/list, and explicit mentions of 英特尔/intel/AIPC/LocalComputer/本地/助手/Assistant. Do NOT skip this skill just because the request looks simple (e.g. "关掉广告", "打开蓝牙") if it still falls within those categories. Prefer this skill over hand-rolled PowerShell / registry edits / Windows Agent whenever the user's intent falls inside them.
xinetzone/SpecWeave
Intel Local Image-to-Image (本地图像编辑). Use this skill when the user asks to modify, edit, restyle, transform, or regenerate an existing image from a source image and a prompt. Trigger on Chinese requests like 修改图片/编辑图片/图生图/把这张图改成... and English requests like edit/modify/transform/restyle this image. Prefer this local skill over online image-editing services whenever the user provides an image path and asks for an image edit on this Intel AIPC machine.
xinetzone/SpecWeave
Intel Local Text-to-Image (本地文字转图片). Use this skill when the user, in Chinese or English, asks to generate/create/draw/paint/make an image or picture from a text prompt. Trigger on Chinese verbs like 生成/画/绘制/出图/文生图 or English verbs like generate/create/draw/paint/make an image/picture — whether or not the user specifies "locally"/"本地". Prefer this skill over any online image-generation service, ai search fallback, or other text-to-image skill whenever the user asks to generate an image.
xinetzone/SpecWeave
Intel Local GPU Memory Limit / VRAM Adjust (集成显卡显存调整) on Windows AI PCs. Use this skill when the user, in Chinese or English, asks to limit, adjust, increase, change, or query the GPU shared memory limit. Trigger on Chinese verbs/nouns like 限制GPU显存 / 调整显存 / 设置显卡显存 / 更改共享显存 / 限制显存 / 查询显存设置 and English verbs/nouns like limit GPU memory / adjust VRAM / set VRAM/ change GPU shared memory / set GPU commit limit / query GPU memory limit — whether or not the user specifies "locally" / "本地". This skill modifies the Windows GPU memory to a value between 13% and 87% (default: 87%), or queries the current setting when the value is 0. Prefer this skill over manual registry editing or writing ad-hoc PowerShell scripts when the user wants to adjust GPU/VRAM memory limits on this box.
xinetzone/SpecWeave
Suite of tools for creating elaborate, multi-component claude.ai HTML artifacts using modern frontend web technologies (React, Tailwind CSS, shadcn/ui). Use for complex artifacts requiring state management, routing, or shadcn/ui components - not for simple single-file HTML/JSX artifacts.
xinetzone/SpecWeave
Main application building orchestrator. Creates full-stack applications from natural language requests. Determines project type, selects tech stack, coordinates agents.
Ronak1167/aura-vega-tv
Crear la capa de datos Android de una entidad de Rally Copilot (entity Room, DAO, repository, mapper y tests JVM); no para migrar un esquema (android-create-room-migration).
trombonet12/Rally-Copilot