标签：#rce

Spent the week breathing new life into DevNotion—59 commits and over 10,000 lines of code later,...

The easiest way to misunderstand LangGraph is to see it as “LangChain, but with more steps.” That misses the point. LangGraph becomes useful when an agent is no longer a single prompt or a simple chain. It becomes useful when the workflow has state, branches, tool calls, human approval, checkpointing, and recovery behavior that must be inspected before the agent is trusted inside a real AI host. I used the Doramagic LangGraph manual as the source-backed reading layer for this note: https://doramagic.ai/en/projects/langgraph/manual/ This is an independent project guide, not an official LangGraph document. I use it as a pre-adoption checklist: what should be understood before wiring a project into Claude, ChatGPT, Cursor, Codex, or another AI host. The point is not to create another prompt library. The useful artifact is a capability resource pack: a manual, source map, boundary notes, pitfall log, smoke check, lightweight eval criteria, feedback notes, and host-ready context that help a developer decide what to verify before adoption. 1. The real boundary is State, not the prompt For a one-shot model call, the prompt is often the main boundary. For LangGraph, the first boundary is the State schema: which fields move between nodes; which fields a node may update; how concurrent branches merge values; which values enter a checkpoint; which values should never be persisted. This is why reducers matter. A message list is usually not just overwritten. It needs an append or merge rule such as add_messages or the TypeScript equivalent. That small implementation detail decides whether parallel work preserves context or silently drops it. My preferred first run is not a “universal agent.” It is a tiny graph with one State schema, one node, one partial update, and one explicit reducer. If that is not clear, adding tools will only hide the problem. 2. compile() is the boundary between description and runtime Before compile() , a LangGraph graph is a description: nodes, edges, c

Last night I ran external security scans on the public websites of 10 leading Shopify and Shopify Plus agencies — the same scan any browser or attacker would see. No credentials, no special access. One agency scored an A. Three scored C- or below. The most common finding appeared on 9 of 10 sites. TL;DR 1 agency scored an A. 3 scored C- or below. 1 scored a D. The most common finding — missing security headers — appeared on 9 of 10 sites. 6 of 10 agencies have no HSTS at all. One agency has a session cookie without the Secure flag. That is the most concrete finding in the set. What was scanned Five categories per domain: TLS (HSTS presence and max-age), security headers (CSP, X-Frame-Options, X-Content-Type-Options, Referrer-Policy, Permissions-Policy), cookie flags, DNS hardening (DNSSEC and CAA) and sensitive exposure paths. All scans run on 23 June 2026. This covers the agencies' own marketing sites — not the client stores they build. Results Agency Domain Score Grade 1Digital Agency 1digitalagency.com 94 A Acidgreen acidgreen.com.au 77 B 30 Acres 30acres.com.au 76 B Fourmeta fourmeta.com 76 B Blend Commerce blendcommerce.com 76 B Elkfox elkfox.com 76 B Charle Agency charleagency.com 62 C Fyresite fyresite.com 62 C Eastside Co eastsideco.com 58 C- Swanky Agency swankyagency.com 55 C- Blubolt blubolt.com 54 D Per-agency notes 1Digital Agency — A (94) HSTS at two years, X-Content-Type-Options and Referrer-Policy set correctly, Permissions-Policy restricting camera, microphone and geolocation, CSP frame-ancestors in place of X-Frame-Options. Only gap is HSTS missing includeSubDomains. Acidgreen — B (77) HSTS with two-year max-age, includeSubDomains and preload — the strongest TLS config in the set. But CSP, X-Frame-Options, X-Content-Type-Options, Referrer-Policy and Permissions-Policy are all absent. Worth noting Acidgreen is multi-platform (Shopify Plus, Adobe Commerce, Magento) rather than Shopify-only. 30 Acres — B (76) A Shopify Plus Partner agency based in Byr

I kept running eslint and thinking my codebase was fine — then someone opened a PR and the first comment was "this function needs JSDoc." The problem: linters check your syntax . Nobody checks whether your exported API is actually documented. Those are two very different things. So I built jsdocscan — a zero-dependency CLI that walks your JS/TS files and flags every exported function or class that is missing JSDoc, or has undocumented parameters. What it catches Errors — exported function or class with no /** … */ block at all: ✗ src/api.js:12 fetchUser missing JSDoc ✗ src/utils.js:34 formatDate missing JSDoc Warnings — JSDoc exists but a parameter has no @param tag: ! src/render.js:7 renderCard undocumented params: opts Exit codes are 0 (all clean), 1 (issues found), 2 (usage error) — pipe-friendly. How to use it # scan a directory npx jsdocscan src/ # Python version pip install jsdocscan jsdocscan src/ # skip @param checks — just verify JSDoc exists npx jsdocscan --no-params src/ # machine-readable output npx jsdocscan --json src/ | jq '.[].findings' # summary line only npx jsdocscan --quiet src/ # custom extensions npx jsdocscan --ext .js,.ts src/ What it skips (intentionally) Non-exported functions and internal helpers — these are implementation details Destructured params ({ a, b }) — too many valid @param opts patterns TypeScript type annotations on params — name: string is stripped, @param name is still required Zero dependencies No parsers, no AST, no require("typescript") . It uses a line-by-line scanner that: Detects export function/const/class patterns via regexes Walks backwards to find a preceding /** */ block Compares @param names in the JSDoc against the actual parameter list npx jsdocscan works with zero install time. pip install jsdocscan brings in nothing extra. Links npm: npmjs.com/package/jsdocscan PyPI: pypi.org/project/jsdocscan GitHub (Node): jjdoor/jsdocscan GitHub (Python): jjdoor/jsdocscan-py Both versions pass the same 38-test suite. Same

I spent the last few months trying to understand why reconciliation errors keep appearing in high-throughput pipelines. Here is what I found. In the race to process millions of transactions daily, modern fintech ecosystems have achieved a genuine miracle of scale. But beneath the surface of that velocity lies a structural problem most engineering teams aren't measuring: causality violations in async event pipelines. Most teams assume that if a transaction shows "Success" in the database, the job is done. At high concurrency levels, that assumption breaks quietly. When "Eventual Consistency" Becomes "Eventual Loss" In distributed systems, Kafka partitions and database shards experience micro-millisecond timing gaps. When a network retry delays a validation webhook, the downstream ledger can commit a wallet update before the validation that should have preceded it completes. To the user, the app glitches. To the engineering team, it's a reconciliation ticket. To the CFO, it's untracked operational cost. The Reconciliation Tax I built a simulation modelling this exact failure mode across 5,000 concurrent transactions. With an 8% network retry probability, conservative for high-traffic payment rails, the causality violation rate was 8.3%. At one million daily transactions, that's over 80,000 unvalidated commits every day requiring manual review. The operational cost compounds across three dimensions: engineering hours spent patching database state, fraud model accuracy degrading on out-of-order training data, and audit trails that cannot demonstrate strict causal sequence to regulators. The Fix The solution is enforcing strict event ordering at the ingestion layer before state commits happen, not better monitoring after the fact. When safeguards including partition-aware routing, exponential backoff, and idempotency controls were added to the same simulation, the violation rate dropped to 0%. Full simulation code and methodology: github.com/yakuburoseline1-gif/cif-simul

I used Postman for years. It was the first thing I installed on every new laptop, the default answer...

OpenAI is attempting to tackle the security issues of the open source software community.

去年我开始给 Hermes Agent 搞知识库。最初的想法很简单：装个向量数据库，查资料时能命中就行。结果跑了一圈发现—— 纯向量检索在 Agent 场景下不够用 。 有些记忆是你能精确描述的（"帮我查上次那篇讲 gbrain 孤页的文章"），有些记忆你只记得个大概（"有个关于知识图谱的……"），有些记忆你甚至不知道你该知道什么（"我有哪些笔记是没关联起来的？"）。一个检索策略覆盖不了这三种场景。 这就是 Knowledge-and-Memory-Management v0.0.2 解决的问题。 三层不是为了「多」，而是为了互补 项目是目前运行在 Hermes Agent 生态里的扩展层，底座是 hermes-memory-installer 提供的 gbrain（知识图谱）+ Hindsight（向量记忆）。KMM 在这上面加了四组模块：采集、笔记/RAG、云盘同步、知识增广。 其中最有工程价值的设计是 三层跨层召回 ： FTS5 全文搜索 → 精确命中，知道你要什么 Hindsight 向量语义 → 模糊配匹，关键词不够时靠语义 gbrain 知识图谱 → 关系探索，你不知道但相关的节点 lightweight_recall.py 把这三级串联成一条管线——先查本地 SQLite FTS5，命中直接返回；没命中走 Hindsight 的文本嵌入匹配；还不满意就用 gbrain 展开知识图谱的邻接节点。 代码在 $AGENT_HOME/scripts/lightweight_recall.py ，调用并不复杂： # 三层召回示例 from knowledge_collector.knowledge_management import KnowledgeManager km = KnowledgeManager () # 一次调用，三级自动回退 results = km . tiered_search ( query = " Agent 记忆系统设计 " , tiers = [ " fts5 " , " hindsight " , " gbrain " ], limit_per_tier = 5 ) for tier , hits in results . items (): print ( f " [ { tier } ] { len ( hits ) } 条结果 " ) for h in hits [: 2 ]: print ( f " → { h [ ' title ' ] } ( { h [ ' relevance ' ] : . 2 f } ) " ) FTS5 走 SQLite 内置，零依赖；Hindsight 端口 8890；gbrain 端口 8787。每层都是独立的可降级——向量服务挂了，全文搜索照样能跑。 真正干活的是这 40+ 工具 三层召回是检索层，采集层才是让知识库有东西可喂的入口。项目塞了 40+ 采集分析工具，按类型分： 网页采集（9 引擎） ：Scrapling 做反检测，Crawl4AI 做智能路由，爬公众号不走弯路 视频采集（12 工具） ：yt-dlp 拖流 → Whisper ASR 转写 → PaddleOCR 关键帧，抖音/YouTube 一把抓 文档分析（9 工具） ：SenseNova 三件套（PDF/PPT/Word）处理扫描件和复杂排版 书籍精炼（6 工具） ： book_to_skill 管线把 PDF/EPUB 拆成结构化 Skill + KMM 笔记 每周日凌晨还有个 knowledge_discovery.py 爬起来，扫 OneDrive 上的新笔记，自动录入 gbrain，顺便跑孤页链接修复——这些属于「你不需要知道它存在，但它让系统不腐烂」的基建。 几个工程取舍 不做纯向量优先 ：FTS5 第一级是因为延迟低（毫秒级），且精确关键词匹配在查代码、查配置时远好于语义 gbrain 孤页处理是刚需 ：知识图谱不加链接维护，三个月后 90% 的页面都成孤岛（亲测 10,666 页从 1,716 个孤页降到 33） 采集优先于检索 ：没有好的采集管线，检索再强也是对着空库跑——项目把采集工具堆到 40+ 不是炫技 适用场景 如果你的 Agent 需要从多个来源（网页、视频、文档、公众号）持续摄入知识，并且知识需要在精确搜索、模糊回忆和关系探索三个维度上都能命中——三层模式值得一试。不需要全部部署，只接 FTS5 + gbrain 两层也能跑，每层独立可降级。 GitHub: mage0535/Knowledge-and-Memory-Management ，MIT 协议。

GitHub热门项目 | The open-source CapCut alternative | Stars: 58,822 | 3,097 stars this week | 语言: TypeScript

I spent a week building Wingman , an open source MCP server that renders a persistent task panel inline in Claude conversations using MCP Apps (SEP-1865). The spec is solid. The SDK is solid. But I hit two bugs that cost me most of a weekend each, and neither is documented anywhere I could find. Writing them up here in case they save someone else the time. Bug 1: resourceUri has two valid-looking locations, and only one works MCP Apps needs a way to tell the host "render this resource as a UI for this tool call." That pointer lives in _meta.ui.resourceUri . The question is: meta on what? I started with a parameterized resource template, ui://wingman/panel/{plan_name} , registered per plan. That was my first mistake. Parameterized templates get listed under resources/templates/list , not resources/list , and hosts do not prefetch or render anything from the templates list. The fix was straightforward once I found it: register one static resource, ui://wingman/panel , and pass the actual plan data through structuredContent on the tool result instead of baking it into the URI. That fix surfaced the real bug. My show_plan tool was returning a plain Python dict: return { " plan " : plan_data , " _meta " : { " ui " : { " resourceUri " : " ui://wingman/panel " }} } This looks correct. It is not. FastMCP's result conversion takes a returned dict and serializes the whole thing into structuredContent , verbatim, including any _meta key the dict happens to carry. So the actual wire result looked like this: result . structuredContent [ " _meta " ][ " ui " ][ " resourceUri " ] # == "ui://wingman/panel", but wrong place result . meta # None — this is what the host actually reads MCP Apps hosts read resourceUri off the top-level _meta on the CallToolResult , not off whatever ended up inside structuredContent . With that pointer effectively missing, the host had nowhere to bind the iframe. The visible symptom was strange: actions in the UI would update on screen but nothing persist

Every time I added AI to a React app, I rewrote the same 200+ lines. Streaming loop. Manual message history. Tool call orchestration. Error handling. setIsLoading(false) only if I remembered. After the third project I stopped and asked: why is nobody solving this the way RTK Query solved REST APIs? So I built Strand ( https://github.com/strand-js/strand ). Before const [messages, setMessages] = useState([]) const [isLoading, setIsLoading] = useState(false) async function send(text) { setIsLoading(true) manually stream tokens manually detect tool calls manually loop until done setIsLoading(false) only if you remembered } After const { messages, send, isPending, isStreaming, cancel } = useConversation({ system: 'You are a helpful assistant.', }) Streaming, history, tool calls, cancellation, retry; all handled. The thing nobody else has: useToolCall Works from ANY component; no prop drilling function WeatherStatus() { const { status, input, output } = useToolCall('get_weather') if (status === 'running') return <div>Checking {input?.location}…</div> if (status === 'done') return <div>{output?.temp}°F</div> return null } Live tool state: pending → running → done. Its observable anywhere in your tree. Fixing the isLoading design flaw The Vercel AI SDK has 4+ open issues ( https://github.com/vercel/ai/issues ) about isLoading getting stuck. The reason is architectural because "request sent" and "tokens arriving" are different states. Strand tracks four: const { isPending, isStreaming, isDone, error } = useConversation() // isPending: waiting for first token // isStreaming: tokens arriving // isDone: just completed // error: something failed Works with Anthropic, OpenAI, and Google Gemini npm install @strand-js/core @strand-js/react zod npm install @strand-js/anthropic # or openai, or google Swap providers by changing one server import. Zero frontend changes. v0.1.8, MIT, open source. → https://github.com/strand-js/strand

"It runs on my own GPU, so it's basically free." I believed that until I put a meter on it. So I ran a controlled benchmark on one box — an openSUSE machine with a single RTX 3090 — driving three local models through ollama under an identical fixed workload (256-token generations in a loop for ~4 minutes each), while my open-source dashboard priced every run by the real GPU energy it burned : power sampled from nvidia-smi every 10 s, integrated over each run's exact window, multiplied by my actual day/night tariff. One number per model, in euros per million output tokens. Here's the part that made me re-run it. The tiny gemma3:1b came out at €0.118 / 1M tokens — about 5× cheaper than a hosted Flash-class API (~€0.55). But gemma3:27b 's electricity alone was €0.706 / 1M — more expensive per token than just paying the cloud, and that's before a single cent of the GPU's purchase price. "Local" didn't make it cheaper; it made it cost more and I own the depreciation. The mechanism is one line: each token costs watts ÷ throughput , and a big dense model is both slow and thirsty. A newer mid-size architecture ( gemma4:26b ) bought a lot of that back, landing at €0.272 . The full guide is methodology-first and reproducible end to end — minting an ingest key, the stdlib-only client, the exact ollama loop that reads eval_count / eval_duration for real tokens-per-second, reading each run back priced, and the honest caveats (this is marginal GPU energy only — not capex, idle, or cooling — and the absolute numbers round to fractions of a cent; the shape is the finding). Read the full guide on Medium → https://medium.com/@arsen.apostolov/how-much-does-it-actually-cost-to-run-a-local-llm-per-million-tokens-measured-4a90a7f31a48

Learn about the progress we’ve made toward our accessibility goals and how you can help make open source more inclusive. The post From pledge to practice: Building a more inclusive open source ecosystem appeared first on The GitHub Blog .

Has anyone else ended up maintaining the same AI skill in multiple places? I use Claude Code, Codex, Cursor, Gemini CLI, Kimi, and several other AI tools. Over time, I accumulated a huge collection of skills and workflows. The annoying part wasn't creating them. It was keeping them synchronized. A skill would exist in one format for Claude Code, another for Cursor, another for Gemini, and so on. Eventually, I got tired of duplicating everything and built an open-source project called AI Omni Skills. The idea is to keep a single source of truth and generate the formats required by different AI tools. Now I update a skill once and regenerate whatever structure a specific tool expects. I'm curious: How are you managing skills today? Are you duplicating them across tools? What integrations would you want to see? Repo: https://github.com/moatazhamada/ai-omni-skills

Cross-posted from the LLMKube blog . A local 27B coding model, running on hardware in my house, is a coin flip. Some runs it nails the fix in twenty minutes. Some runs it edits the wrong file, writes a test that passes no matter what the code does, and tells you it's done. The bet behind LLMKube's Foreman was never that I would find a local model good enough to trust. It was that I could build a harness I trust more than any single model's output. This weekend tested that bet harder than any benchmark could, because the harness spent the weekend building its own guardrails. Here is the short version of what happened across 0.8.12 and 0.8.13. My local coder built three new gates for itself. One of them shipped with the exact flaw it was written to catch, and the review caught it. Three new contributors sent four clean pull requests while the machines worked. The same model ran on an AMD box and an Apple Silicon Mac, and the Mac quietly won a round nobody expected. And not one byte of any of it touched a cloud API. The thesis, stated plainly Trust the harness, not the model. A coding agent on a local model produces output of wildly variable quality, and no amount of prompt tuning makes a 27B as reliable as a frontier model. So Foreman does not ask the model to be reliable. It wraps the model in a pipeline that is : the coder works in a cloned workspace, a fast in-workspace gate runs gofmt, vet, build, lint, and the unit tests for the packages it touched; a reviewer reads the diff against the issue; and a clean-room Kubernetes Job re-runs the full suite before anything is allowed to call itself a GO. Around all of that sit deterministic rails: scope checks, edit-free-streak detection, repo-map context. The model is a stochastic component inside a system whose job is to make the system's verdict trustworthy even when the component is not. The interesting question is never "is the model good." It is "does the harness catch the model when it is bad." This weekend gave me

GitHub热门项目 | Scalable datastore for metrics, events, and real-time analytics | Stars: 31,581 | 30 stars this week | 语言: Rust

GitHub热门项目 | Material UI: Comprehensive React component library that implements Google's Material Design. Free forever. | Stars: 98,453 | 60 stars this week | 语言: JavaScript

GitHub热门项目 | 🔰 Home Assistant Operating System | Stars: 7,212 | 37 stars this week | 语言: Python

GitHub热门项目 | A portable accelerated SQL query, search, and LLM-inference engine, written in Rust, for data-grounded AI apps and agents. | Stars: 2,981 | 0 stars today | 语言: Rust

GitHub热门项目 | Tunnel all your traffic over Websocket or HTTP2 - Bypass firewalls/DPI - Static binary available | Stars: 6,841 | 24 stars today | 语言: Rust