GitHub AI Daily

⭐⭐⭐ "Skills" today showed up in three institutionalized shapes at once: vendor-maintained, community-maintained and pape…

• Browser vendors start giving "web platform knowledge" an evaluation dashboard: Google Chrome turns 203 guides into an asset that needs continuous regression testing
• Clinical knowledge gets an official skill-library shape for the first time: CQF turns CQL, FHIR and clinical reasoning into Agent Skills that ship with evals
• An evidence-grounded skill library reaches industrial scale: ASB distils 11,830 metabolomics skills out of methods papers

The single most important thing today is that evaluators themselves now have to be evaluated. The ultrasound report meta…

• Evaluating the evaluators: BLEU and BERTScore cannot catch planted clinical errors — structured fact extraction plus image localization can
• Skills need quality gates too: NVIDIA now separates a failed evaluation arm from an arm that never ran
• Label photos into the app: food-label recognition is split into "AI reads only, values normalized per 100 g, energy cross-checked against macros"

⭐⭐⭐ Skills are turning from "prompt fragments" into portable, evaluable, updatable methodology assets - and a public ind…

• awslabs/hcls-agent-skills: AWS turns 43 healthcare and life-science skills into portable standard parts, and proves with 410 pairwise prompts that they beat going without
• openai/codex-security: the official agent security scanner, now starting to pull other vendors' findings into a single evidence chain
• awslabs/unified-kg-rag-on-aws: four graph RAG methods on one shared index - and renaming an entity is now treated as a migration that can be rolled back

⭐⭐⭐ "Did you actually read it" is becoming a question a machine can fail. On 11-10 I noted astryx reporting named scenar…

• Prabh54/groundedmed: A biomedical RAG agent that has to prove it did not make anything up
• Muratovnik/assay: A gate that forces coding agents to list every file they actually read
• vxxqv/openomicsbench: A locally runnable answer to "how do you validate omics software?"

The clearest thread in this issue is that AI systems are now being required to be independently verifiable, and this iss…

• meta/astryx: Meta open-sources the design system it used internally for eight years, with 150+ components and fully open internals — the problem it solves is not looking good, but letting an agent read and modify a UI
• github/squad: GitHub itself ships a "human-led AI development team" — it puts each team member's charter (job description) into repo files, so governance and code live in the same git
• wld007/dsh-medical-anonymizer: a "bulk medical-record de-identification" plugin for DeepSeek Harness — rules run first, the model reviews blind, and anything unresolved is never silently accepted, with an audit report for every batch

The clearest thread in this issue is that AI-generated artefacts are now systematically required to substantiate themsel…

• open-mtor-atlas/atlas: an mTOR research database graded by what the evidence can carry — 400+ primary studies where every claim names its source, its measurement conditions, and the point where it stops holding
• hherb/bmlib: turning "can this paper be trusted" into programmable infrastructure — a four-state retraction verdict, Cochrane nine-domain risk of bias, and a PICO where every part is backed by the paper's own quoted passages
• ccch1mneyyy/dsh-Go: putting DeepSeek Harness in your pocket — it runs locally on the phone and can remote-control your computer, and it lets you talk over the agent mid-answer

The clearest thread this issue is that constraints are sinking from "requirements written in a prompt" down into "gates …

• nicolaguglielmi/lights-out: turning "let AI write the code" into a lights-out factory — eight numeric gates judge every PR, every ticket gets its own perimeter, a bad canary rolls itself back, and the last step always needs one human GO
• Rul1an/assay: an offline-recomputable evidence profile for privileged tool calls — it states plainly that "a tool returning success is the provider's assertion, never proof", and grades claims into four tiers
• rootstudioyaml/sprag: turning "which model should do this job" into a recomputable rule — the same 12 requests cost $1.83 instead of $6.68, and every rule comes with a replay case you can open

The most visible thread this issue is that constraints are moving from "requirements written in a prompt" to "gates that…

• nicobailon/skill-diet: putting an agent's skill pack on a diet — 70–80% of the tokens can be cut with no drop in test results, and it turns "does this still work on other models?" into a recomputable table
• qf-studio/navigator: moving context management out of the prompt and into the runtime — it blocks the exit signal when a task isn't finished, blocks hand-written task files, and logs one line for every rejection
• Mason-Levyy/myfitnesspal-mcp: letting AI write straight into your food diary — but every write must be verified by reading the diary back, the simplest and most reliable anti-hallucination technique I have seen

The clearest thread this issue is that "write the promise into code, not into the prompt" is spreading from medical AI i…

• santiagopatinoserna/AdelIA: moving the "clinical promise" out of the prompt and into code — every clinical statement must carry a citation, no evidence means a full refusal, and red-flag signals short-circuit the model entirely
• 12burb/hermes-fledge: taking the job of grading agent-written skills away from the model — skills have to earn their wings on evidence, and a skill born from a bad run goes on probation instead of shipping
• senseiewok/cf-skills: a cystic fibrosis research lab turned its evidence practice into 47 commands — every answer is an evidence record with source, date, the exact fields relied on and a "limitation the code refuses to leave empty", and it never calls a language model

The judgement most worth carrying out of this issue: the competitive focus in AI tooling is shifting from "who can gener…

• ovumcy/ovumcy-web: turning "can this prediction be trusted" into a gate in code — it does not merely compute the fertile window, it withholds the whole thing until there is enough data
• vercel-labs/agent-desktop: letting an agent operate the desktop without "guessing coordinates from a screenshot" — it reads the OS accessibility tree and hands every actionable element a stable reference, and today it added MCP
• zarnautovic/tdeemon: writing the hard floors of a calorie or supplement plan into code — the AI may propose but never execute, and there is a floor nobody can lower

Of the 134 entries today, the most notable thing is not which repository gained how many stars, but that "evidence" and …

• ClawBio/ClawBio: writing a domain expert's analytical decisions into a versionable, reviewable, reproducible machine contract — the way this skill library is organised is precisely the layer my GRPM nutrigenomics system is missing
• cortexkit/aft: giving an agent a real IDE and OS — swapping "read the file, change the code" for "read and write by symbol", and keeping a restorable backup behind every change
• zilet/cairn: reading labs, training, food and sleep as one picture, then offering only the one thing worth doing today — and "no scores, no nudges" is its most counter-cultural sentence

Across the 102 entries today there is a single through-line: the agent ecosystem is starting to show layered self-doubt …

• Alice39s/agent-stash: a "skill drawer" for DSH — it distils reusable workflows out of finished sessions and saves them as Skills after review, which is exactly the missing piece in my expert-package workflow
• MarissaFamularo/Calibrate: turning "should the doctor trust the AI" into a measurable 2×2 matrix — over-reliance and reflexive dismissal name the exact risk I carry when producing supplement plans
• richmondgoh8/macro-longevity: writing a "coverage regression" into a health app — 140 lab references each mapped to a usable field, the single most transferable piece for the 4P supplement platform

The single trend most worth recording today is that the centre of gravity in the AI agent ecosystem is shifting from "ca…

• can1357/oh-my-pi: the IDE is now genuinely wired into the coding agent — the 37,251-star pi fork ran 74 commits in 24 hours, and today's theme is "make the agent guess less, measure more"
• ayuayue/PiDeck: the Chinese community packs both pi and DSH backends into one desktop workbench — and DSH is the very harness running on my machine
• OpenAdaptAI/OpenAdapt: turning "the agent says it succeeded" into a falsifiable claim — on an open-source EHR, 30 deliberately corrupted records were caught by read-back verification

The most notable thing today is that "credentials must not enter the model's context" moved to a layer earlier than runt…

• lightpanda-io/browser: The AI browser starts hiding passwords from the model — scrubbing credentials out of the context the model sees is one layer earlier than any runtime interception
• seancpc/MedAgent-EHR: Planner–Executor–Verifier plus two-phase writes put a clinical agent at ~88% on 300 MedAgentBench tasks, running entirely on a self-hosted local model
• whiteravens20/diet-app: Nutrition math becomes a deterministic engine and AI drops to optional assistant — "it never invents nutrition facts" is a single sentence in the README

The most memorable item today is Tencent Cloud formally establishing "cloud development backend" as an agent ecosystem s…

• TencentCloudBase/CloudBase-AI-Toolkit: Tencent turns "cloud development backend" into a plugin, skill and MCP triple-decker for AI coding agents
• itsoso/health-llm-driven: The Chinese-language community ships a real product for the "personal health OS" — digital health twin, safety guardian, action ranking, intervention cycle
• craft-b/formula-forge: A clinical nutrition formulation agent draws a hard boundary — the model may only propose structure, all nutrition numbers are recomputed by the deterministic layer

The clearest theme today is that "local-first" has stopped being a slogan and become an architectural constraint — one t…

• cindy: "Works out of the box" AI Agent built as a complete product across desktop, mobile and group chat — not yet another terminal framework
• open-healthcare-agentic-platform: medical agents start answering head-on that "the model doesn't get the final say — the rules do"
• stjude/MyGPT: clinical-grade RAG that "runs on a $1,839 Mac mini," with the paper and the evaluation code open-sourced together

The most consistent action today is not "build a stronger agent" but "give an agent a cheap dispatch layer". mu hands ro…

• mu (μ) gives a coding agent a small, fast judge — and lifts routine decisions off the big model
• bmlibrarian is a multi-agent biomedical literature system — and today it fixed a hard bug: reading study counts as sample sizes
• OpenNutriTracker v2.5.0: an offline-first calorie and micronutrient tracker, with a new "to-log list" capture flow landing on Oct 24

Write permission has become the dividing line for clinical agents. Yesterday's clinical agents were still competing on "…

• Omnigent: an 11k-star "meta-harness" that pulls Claude Code, Codex and Cursor under one roof — and starts governing them
• Last EHR turns "an agent writing to the chart" into an approvable protocol — and publishes its own coverage ceiling
• An HL7 official repository grows a molecular tumour board from patient banner to AI synthesis in one day

Differentiated bets after the harness race went white-hot: hermes-agent (hibernating environments + six backends), openh…

• Vector Institute open-sources Odyssey: an interpretable patient-timeline foundation model that treats "every claim is falsifiable" as engineering discipline
• Google releases Elephant-Goldfish Model: a spec-driven development plugin where zero-context subagents stress-test your spec before you write code
• Nous Research Hermes Agent: the 256k★ open-source agent harness that makes "environments sleep when idle" a default capability

The bottleneck has moved from the model to the collaboration protocol. The three hottest AI coding projects today — git-…

• git-ai — an open replication of Anthropic's internal Scion, "source control redesigned for agents"
• Tencent Hunyuan Hy-Memory — coding memory turned into a standalone dual-system cognitive engine (MemHub connects five coding tools)
• HasteHealth — an open-source FHIR server for AI-native health apps, with MCP and SMART on FHIR built in

The skill ecosystem enters supply-chain governance. NVIDIA defines "verified skills" with signatures, evaluation sets an…

• NVIDIA/skills — skills now have signatures, evaluation sets and benchmark reports
• jcode — the real bottleneck in parallel agents is handoff, not start-up
• NotePilot v1.8 — the point of a clinical summary is being verifiable, not sounding right

Medical agents compete on the middle layer, not on capability. MedMCP keeps tools and data inside the hospital, HealthCl…

• MedMCP — medical imaging agents choose the "inside the data centre" route
• M4 v0.5.9 — the clinical semantic layer starts versioning skills like software
• ai-memory — coding-agent memory is becoming cross-vendor infrastructure

Official skill supply appears. android/skills is the first time a platform vendor defines agent skills from its own repo…

• MadsLorentzen/ai-job-search — the agent's first white-collar workflow hit
• android/skills — a platform vendor starts defining "skills" itself
• Aperivue/medsci-skills — medical skill libraries go from broad to deep

Skills are becoming personalities. mattpocock/skills, agency-agents and ponytail together show agent skills evolving fro…

• deepseek-ai/deepseek-harness — a model vendor builds a plug-in agent runtime itself
• mattpocock/skills — 156k stars prove "methodology beats model"
• MedicalHarness — a quarter of a medical agent's score belongs to the harness

The agent-interface debate has split into camps: MCP (standardised, but humans build and maintain each server) meets CLI…

• CLI-Anything — the other road besides MCP: translate the command line into an agent interface
• Medical evidence synthesis becomes auditable — ReviewPilot runs PRISMA as a multi-agent pipeline with provenance
• One person plus one coding agent open-sources a complete MRI toolkit

The skill becomes the unit of distribution. In both coding and medicine, competition has shifted from models to "skill p…

• Medical AI skill libraries reach "thousand-item" scale — OpenClaw-Medical-Skills and scientific-agent-skills land the same day
• On-device clinical de-identification — OpenMed makes HIPAA compliance a local capability
• Food and nutrition AI moves from "logging" to "auditing" — FoodTruth uses a vision model to break packaging claims