“arbor”的搜索结果

Autonomously improve a real artifact (code, training recipe, agent harness, data pipeline, prompt) against an objective and an evaluator, using Hypothesis Tree Refinement (HTR) from the Arbor paper. Use this whenever someone wants to iteratively optimize something over many experiments without overfitting — e.g. "get my model's eval score up", "improve this agent/harness", "tune this pipeline", "beat the baseline on this benchmark", "run a search over approaches and keep the best", "do an MLE-bench / Kaggle-style optimization", or any long-horizon "make this artifact better and don't just memorize the dev set" task. Trigger it even when the user doesn't say "Arbor" or "hypothesis tree" but describes repeated experiment-and-evaluate loops, branching exploration of competing ideas, or worries about a dev/test gap. Runs Claude itself as the coordinator with subagent executors in isolated git worktrees; for the standalone `arbor` CLI tool see references/arbor-upstream.md.

中文简介正在准备中。当前显示原文说明。

Autonomously improve a real artifact (code, training recipe, agent harness, data pipeline, prompt) against an objective and an evaluator, using Hypothesis Tree Refinement (HTR) from the Arbor paper. Use this whenever someone wants to iteratively optimize something over many experiments without overfitting — e.g. "get my model's eval score up", "improve this agent/harness", "tune this pipeline", "beat the baseline on this benchmark", "run a search over approaches and keep the best", "do an MLE-bench / Kaggle-style optimization", or any long-horizon "make this artifact better and don't just memorize the dev set" task. Trigger it even when the user doesn't say "Arbor" or "hypothesis tree" but describes repeated experiment-and-evaluate loops, branching exploration of competing ideas, or worries about a dev/test gap. Runs Claude itself as the coordinator with subagent executors in isolated git worktrees; for the standalone `arbor` CLI tool see references/arbor-upstream.md.

中文简介正在准备中。当前显示原文说明。

Adversarial reviewer for The Federated Harbor whitepaper — the third paper in the Curiositech sequence after Anchor and Bonded Commons. Probes federation-specific threats: trust transitivity, cross-harbor token forgery, federated revocation under partition, cross-harbor Sybil, cross-domain settlement, equivocation, bond-pool draining, cold-start joining, federation-operator Sybil. Use during a versioned red-vs-white round; pairs with federated-harbor-whitehat. NOT for incident response, NOT for the Anchor or Bonded papers (use redteam-review).

中文简介正在准备中。当前显示原文说明。

Defensive counterpart to federated-harbor-redteam: closes the smells that red team opens against The Federated Harbor whitepaper. Federation-specific defenses for cross-harbor primitives — mechanization commitments, scope hedges, cross-paper dependency formalization, pre-emptive analogies (CT, Macaroons, atomic swaps), and defense-in-depth framing (refuses vs prices). Use during a versioned red-vs-white round; pairs with federated-harbor-redteam. NOT for ad-hoc code review, NOT for the Anchor or Bonded papers (use whitehat-defense).

中文简介正在准备中。当前显示原文说明。

Organize repeated experiments to improve models, prompts, or agent systems against measurable goals. The skill records hypotheses and lessons in a persistent tree, isolates trials, and compares promising candidates using evaluation data kept separate from development.

中文简介正在准备中。当前显示英语简介。

• Compare repeated experiments to improve model training
• Improve an agent system's measured success rate
• Test prompt changes against a fixed evaluator

Hard-gated ideation for an Arbor research run. Load at the START of every IDEATE round, before drafting any hypothesis. Enforces the PI mindset (mechanism over knob), the four-question first-principles probe, the kill-filter, and the four-line hypothesis format. Ported from Arbor's idea_drafting + first_principles_probe.

中文简介正在准备中。当前显示原文说明。

The Arbor executor workflow: clone the repo from the git bundle your brief hands you, implement and evaluate exactly ONE hypothesis with the file tools, then report structured results via worker_complete. Preloaded automatically on arbor-executor task workers. Never touch the held-out test split.

中文简介正在准备中。当前显示原文说明。

Capstone/orchestration skill — build an M-Agent + N-Human cooperative IDE in Rust gpui (the Harbor): many agents and humans co-editing the same files as co-equal CRDT replicas, governed by claims/guard/salvage, across LAN/shared/remote harbors. The INDEX that dispatches into the sibling rust skills. Use when building the collaborative editor, the agent-fleet console, multiplayer editing with agents-as-peers, or any slice of the Harbor. Trigger on: cooperative IDE, collaborative editor, multiplayer editor, agents and humans co-editing, gpui IDE, Loro CRDT editor, harbor editor, claims/salvage, "build the cooperative IDE". NOT for: a single non-collaborative gpui screen (compose the siblings directly), web editors, or non-editor apps.

中文简介正在准备中。当前显示原文说明。

Implements HIPAA de-identification methods under 45 CFR §164.514(a)-(b). Covers expert determination method and safe harbor method with 18 identifiers removal, re-identification risk assessment, limited dataset requirements, and data use agreements. Keywords: HIPAA de-identification, safe harbor, expert determination, 18 identifiers, limited dataset, PHI.

中文简介正在准备中。当前显示原文说明。

Implements Children's Online Privacy Protection Act (COPPA) compliance under 16 CFR Part 312. Covers verifiable parental consent methods including signed forms, credit card verification, government ID, knowledge-based authentication, and video call. Includes FTC safe harbor programs and enforcement actions. Keywords: COPPA, FTC, children, parental consent, safe harbor, verifiable consent.

中文简介正在准备中。当前显示原文说明。

DECIDE-phase doctrine for an Arbor research run: when to merge, prune, combine, or finalize; how the held-out merge gate works; and the rule that the final report uses TEST scores. Load before deciding what to do with completed experiments.

中文简介正在准备中。当前显示原文说明。

The Arbor research-coordinator protocol: the OBSERVE -> IDEATE -> SELECT -> DISPATCH -> DECIDE cycle over a durable Idea Tree. Preloaded automatically on /auto-research sessions. The coordinator never edits code or runs commands — executors do that in isolated worktrees; the coordinator steers the tree with idea_tree / dispatch_experiments / merge_experiment.

中文简介正在准备中。当前显示原文说明。

Set up and run local LLMs with Harbor. Use when the user wants to run models locally, install Harbor, start Open WebUI, llama.cpp, Ollama, vLLM, Docker Model Runner, MLX, or oMLX, pull GGUF or HuggingFace models, add SearXNG web search, Speaches TTS/STT, or Open Terminal code execution, launch Codex/Claude/Grok/OpenCode against a Harbor backend, or troubleshoot GPU, VRAM, and service startup.

中文简介正在准备中。当前显示原文说明。

Harbor is an open-source container registry that provides security features including vulnerability scanning (integrated Trivy), image signing (Notary/Cosign), RBAC, content trust policies, replicatio

中文简介正在准备中。当前显示原文说明。

Rank possible gene regulatory relationships from bulk or single-cell expression data using GRNBoost2 or GENIE3. The skill guides input preparation, compatible execution, method comparison, and stability checks to help prioritize candidates for further biological investigation.

中文简介正在准备中。当前显示英语简介。

• Rank candidate regulator–target gene relationships
• Compare GRNBoost2 and GENIE3 results
• Check network stability across seeds and resampled data

Review incoming GitHub issues and pull requests, verify claims, and organize evidence alongside existing decisions. Prepare unresolved choices with recommendations and consequences for maintainers, then carry out actions covered by verified authorization without merging changes.

中文简介正在准备中。当前显示英语简介。

• Sorting incoming bug reports and feature requests
• Applying approved rules to duplicate reports
• Preparing evidence and options for maintainer decisions

Guides migration of historical SWE-Marathon agent configurations to the shared Codex runtime. It preserves native tasks, verification rules, and scoring while outlining environment checks and restrictions on retired execution methods.

中文简介正在准备中。当前显示英语简介。

• Migrating historical SWE-Marathon agent configurations
• Preserving task and scoring rules during migration
• Checking the native environment before a full run

Use when the user hands over a dsh plugin repository (or a real migration commit / version corridor) and wants its upgrade experience extracted into one auto-graded Harbor benchmark exam task — produce tasks/<id>/ (fixture + instruction.md + task.toml + judge.mjs + solution) that harbor run can grade 0–1 and that folds into the dsh-plugin-upgrade-skill benchmark suite; also applies to turning existing version cards (references/v*.md) into executable exam tasks. Not for writing upgrade cards; not for running the benchmark.

中文简介正在准备中。当前显示原文说明。

Inspect an agent repository and optional traces, interview the user, write reviewed Task Specs, build and audit Harbor tasks, and bootstrap reusable project World Knowledge Skills. Use for agent evals, benchmark design, Task generation, controlled Environments, synthetic data, Verifiers, Harbor runs, calibration, or continuous benchmark maintenance.

中文简介正在准备中。当前显示原文说明。

Author, package, validate, or run mathematical evaluations as Jacobian Harbor datasets.

中文简介正在准备中。当前显示原文说明。

Intake for an autonomous research run (Arbor). Use when the user wants to optimize a metric in a repo over many isolated experiments — 'optimize this benchmark', 'improve the model F1 overnight', 'beat the leaderboard'. Discovers the repo/eval/splits, measures the baseline, confirms a Research Contract, then emits a ready-to-send /auto-research command. Does NOT start the run itself.

中文简介正在准备中。当前显示原文说明。

Create custom modules for [Harbor Boost](https://github.com/av/harbor/tree/main/boost), an optimizing LLM proxy. Use when building Python modules that intercept/transform LLM chat completions—reasoning chains, prompt injection, structured outputs, artifacts, or custom workflows. Triggers on requests to create Boost modules, extend LLM behavior via proxy, or implement chat completion middleware.

中文简介正在准备中。当前显示原文说明。

Metal/wgpu/WGSL shader surfaces for native Rust gpui apps (Zed-family, pd-console), stockpiled with beautiful copy-pasteable shader-toy examples. Use for custom GPU fragment passes behind/around gpui panes: ocean/water shaders, pixelated waves and boats, a living harbor, dithered chrome borders, sonar sweeps, aurora/starfields, CRT/scanline post. Trigger on: wgsl, wgpu, metal shader, gpui shader, shadertoy, fragment shader, SDF, noise/fbm, ordered dithering, pixelation, render-to-texture, "pixelated waves and boats", living harbor water. NOT for: web/GLSL/three.js shaders (use a web tool), non-shader gpui motion (use rust-gpui-motion), general GUI layout/color (use beautiful-gui-design), CLI/TUI (use beautiful-cli-design).

中文简介正在准备中。当前显示原文说明。

Expert landscape designer transforming yards through photo mapping, 3D visualization, seasonal planning, and deep plant knowledge. Specializes in fast-growing privacy screens (knows arborvitae pitfalls), architecture-appropriate design, outdoor living spaces, and realistic maintenance expectations. Activate on "landscape design", "yard design", "garden planning", "plant selection", "privacy screen", "outdoor living", "backyard makeover", "arborvitae", "hedge", "fast growing tree", "landscaping ideas". NOT for interior design (use interior-design-expert), hardscape construction (consult contractors), or lawn care chemicals (consult local experts).

中文简介正在准备中。当前显示原文说明。