「decide」の検索結果

Audit whether a paper's baseline comparisons are COMPLETE, FAIR, and SIGNIFICANT: a required recent SOTA baseline is missing while 'best/SOTA' is claimed (HP-MISSING-BASELINE); a baseline is undertuned / given less compute-tuning-data, run at a mismatched config, or the equal-budget ablation-as-baseline is absent (HP-WEAK-BASELINE); 'outperforms' is asserted over overlapping error bars or with no variance/seeds (HP-SIG-OVERLAP); and a cross-row 'improves over baseline by X%' is arithmetically wrong (HP-DELTA-ERROR, cross-row form only). A versioned per-domain baseline profile + a live leaderboard/recency search are assembled by the EXECUTOR as structured facts; a fresh cross-model reviewer (gpt-5.6-sol xhigh, read-only, fresh thread per dimension) PROPOSES findings, each span-anchored to a ledger claim_id; tools/adjudicate_findings.py DECIDES the verdict. Works at L0 (stated comparisons) and deepens at L2 (configs/result files). A completeness question it cannot settle internally becomes needs_external_check, never a guessed missing baseline. Emits baseline-comparison-audit.findings.json; computes NO verdict. Detect-only. Triggers: "baseline audit", "missing baselines", "is the comparison fair", "weak baseline", "baseline 误报", "SOTA earned?".

日本語の概要は準備中です。原文の説明を表示しています。

Audit whether a paper's EVALUATION DESIGN actually measures what it claims and whether its reporting is complete — the validity layer family D (experiment-forensics) cannot reach. Three patterns: train/test leakage means the reported score may not measure generalization (HP-EVAL-LEAKAGE — adopts the Kapoor & Narayanan 8-type / 3-category leakage taxonomy; the illegitimate-proxy / sampling-bias / pretraining-contamination subtypes hand off as needs_external_check, naming but NEVER running Oren-2023 exchangeability / Shi-2023 Min-K% / Golchin-2023 Time-Travel / BIG-bench canary); a load-bearing LLM judge is conflicted (same model/family as a compared system) or unvalidated (no human-agreement, no bias control) (HP-JUDGE-VALIDITY); a declared condition/metric is dropped or switched to favor the method, or 'best' is chosen with no held-out set (HP-SELECTIVE-REPORTING). Verdict-bearing at L0/L1 from the DESCRIBED protocol — NOT repo-gated like experiment-forensics; L2 only CONFIRMS against split/preprocessing/result files. A fresh cross-model reviewer (gpt-5.6-sol xhigh, read-only, fresh thread per pass) PROPOSES findings, each span-anchored to a ledger claim_id; tools/adjudicate_findings.py DECIDES the verdict. Leakage and under-reporting are usually HONEST methodological errors — every finding describes a discrepancy to CHECK, never an accusation. An LLM generating GROUND-TRUTH labels is HP-FAKE-GT (experiment-forensics) — routed there, not here. Emits eval-design-forensics.findings.json; computes NO verdict. Detect-only. Triggers: "eval design audit", "evaluation validity", "train/test leakage", "data leakage", "is the score measuring generalization", "LLM judge bias", "is the judge validated", "selective reporting", "cherry-picked results", "评估设计审计", "评测有效性", "数据泄漏", "训练测试集泄漏", "裁判模型有没有验证", "选择性报告".

日本語の概要は準備中です。原文の説明を表示しています。

Decide which model, effort level, and cascade shape each subagent gets, and how to keep improvement loops safe (evaluator-as-selector, stop on regression). Routes on measured cost-per-completed-task rather than per-token price, because a tier's token count varies more by task shape than price varies across tiers. Covers per-model effort semantics, the concision lever, cascade preconditions, context handoff, and watching a subagent fan-out live. Use when choosing a subagent's model or effort, when choosing how to escalate a failed attempt, or when asked which model or effort a task should get. Grounded in a 2026-07 calibration, a 2026-08 coding-cost study, a 2026-09 seeded-bug repair battery, a 2026-10 Haiku 5.5 ladder priced per spawn, and a 2026-10 retry-versus-escalation ladder on 295 SWE-bench Verified tasks; Managed Agents API specifics are operational, not calibrated.

日本語の概要は準備中です。原文の説明を表示しています。

Decide and implement how to support a new LLM provider or agent engine in AWF - either as a proxied provider (api-proxy adapter) or a direct-API engine (domain allowlist only), e.g. Cursor, Aider, or any tool that calls its own endpoint.

日本語の概要は準備中です。原文の説明を表示しています。

Record user-decided ADRs or inspect their health read-only. Preserve attribution and acceptance evidence.

日本語の概要は準備中です。原文の説明を表示しています。

Record user-decided ADRs or inspect their health read-only. Preserve attribution and acceptance evidence.

日本語の概要は準備中です。原文の説明を表示しています。

Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.

日本語の概要は準備中です。原文の説明を表示しています。

Use when 王建硕 wants to decide whether a VoiceDrop article should get an AI 题图 (公众号封面), and which style of image prompt to use — before generating any cover. Triggers — "这篇要不要配题图", "选个题图风格", "给这篇出个题图 prompt", "choosing cover", "/wjs-voicedrop-choosing-cover".

日本語の概要は準備中です。原文の説明を表示しています。

Draft a reply or a value-add quote tweet for a specific X (Twitter) tweet from its URL. Use to reply in a thread, answer a creator, or quote-tweet with added value. Parses the tweet URL, reads pasted context, decides reply vs quote tweet, and drafts 1-3 variants in your voice. X has no LinkedIn-style 2-level thread flattening, so a reply is just a tweet. Publora has no reply endpoint, so the draft is returned as a copy-paste block to post yourself. Not for top-level posts (use x-post-writer).

日本語の概要は準備中です。原文の説明を表示しています。

Brainstorms and debates approaches, then drives toward an actionable decision. Use whenever someone needs a thinking partner for a decision they're facing: 'discuss', 'debate', 'brainstorm', 'weigh options', 'tradeoffs', 'should I do X or Y', 'help me decide', 'I'm torn between', 'sanity check my thinking', or 'what do you think about'. The user must be asking for help reasoning through a choice — not asking to build, fix, evaluate, plan, or modify something (even if the topic involves this skill itself). Picks the right decision lens, surfaces tradeoffs and blind spots, pushes back when reasoning is genuinely weak, and never implements.

日本語の概要は準備中です。原文の説明を表示しています。

This skill should be used when the user asks to "direct a motion graphics video", "do creative/art direction for a video", "define a motion language", "set the tone and pacing", "make the animation feel consistent", "decide what to animate (and what not to)", "turn a brief/brand into motion choices", or needs the senior creative judgment that sits above specific techniques (gsap/lottie/shader). Covers motion language, tone & energy, pacing/rhythm, motion hierarchy, restraint, and cross-shot consistency.

日本語の概要は準備中です。原文の説明を表示しています。

This skill should be used when the user asks to "compose a frame", "set up a grid", "plan shot composition", "place the focal point", "decide where elements enter and exit", "add a parallax/camera move", "adapt a layout for 16:9 / 9:16 / 1:1", or "keep content inside safe areas". Covers grids, safe areas, focal hierarchy, and 2D/3D camera framing.

日本語の概要は準備中です。原文の説明を表示しています。

Decide whether a preview the visual-diff bot flagged actually regressed or is simply nondeterministic, using a repeat-render oracle at a single commit. Use when a PR's preview diff reports a changed preview whose source the PR does not touch, or when a render, GIF or filmstrip is suspected of being unstable.

日本語の概要は準備中です。原文の説明を表示しています。

Prioritize assumptions by risk and importance. Use after assumption mapping to decide what to test first.

日本語の概要は準備中です。原文の説明を表示しています。

Use when an editor MCP server (unity, godot, unreal, blender) or the browser MCP shows up in your tool list - what each one can observe or change, how to check it is attached to YOUR task workspace, why the repo's tests still decide, the Blender code-execution caution, and the headless fallback when none is connected.

日本語の概要は準備中です。原文の説明を表示しています。

Use when an analiz question needs the human, not the code, to decide - recording it with record_open_questions, judging blocking vs non-blocking, and resuming the report after an answer

日本語の概要は準備中です。原文の説明を表示しています。

Use when starting a mobile task that could be Flutter or native - decide from the existing app, platform-specific needs, and the analiz task

日本語の概要は準備中です。原文の説明を表示しています。

Use when a stakeholder sends a request - route it to backlog tasks, an analiz task, or a focused question, decide ask-vs-assume with a Definition of Ready, and ask only what a wrong guess would actually cost

日本語の概要は準備中です。原文の説明を表示しています。

Use when a release reaches awaiting_verdict (or you are woken for an incident) - how to read runtime logs, error groups, health and smoke evidence for the release's component and decide finish, re-check, roll back or hand to a human

日本語の概要は準備中です。原文の説明を表示しています。

Use when you are about to hand off any change to a screen or component on a mobile repository - the size, text-scale and dark-mode matrix, the exact tool calls to see your own build, and the checklist that decides whether the screen is done.

日本語の概要は準備中です。原文の説明を表示しています。

Use when building Flutter screens - compose small stateless widgets, keep layout declarative, decide layout by available width not device/orientation, separate presentation from business logic, and use the theme not hardcoded styles

日本語の概要は準備中です。原文の説明を表示しています。

Use when you are about to hand off any UI change - the four-width check and the screenshot checklist that decide whether the screen is really done.

日本語の概要は準備中です。原文の説明を表示しています。

Use when starting a backend task that could be built in either language - decide Go vs Java (Quarkus/Spring) from the task's shape and the repository's existing stack

日本語の概要は準備中です。原文の説明を表示しています。

Aurora Smart Home orchestrator — routing layer for all smart home skills. Use this skill when the user asks ANY smart home question and you need to decide which skill to invoke, or when a task spans multiple skills (e.g., "build a sensor that shows on a dashboard and triggers automations"). Invoke aurora FIRST before reaching for a specific skill — it will route to the right specialist(s) and recommend the correct Claude model to keep token usage efficient. Trigger on: smart home, Home Assistant, ESPHome, automation, IoT, dashboard, ESP32, Node-RED, or any request about controlling or monitoring devices at home.

日本語の概要は準備中です。原文の説明を表示しています。