by 安全扫描
OpenClaw
安全
high confidence该技能的代码、指令和要求与其声明的目的相符（SOTA CV/DS 竞赛的规划和工件支架），不请求无关的凭据或静默网络安装。
评估建议
该包看似连贯，专注于竞赛支架，但在运行前：（1）检查脚本以确认 URL/路径/引用清洗行为；（2）在专用竞赛工作空间运行脚本；（3）避免在暴露秘密的环境中运行；（4）如果与真实浏览器/VM 自动化配对，审查执行路径代码以防止意外网络泄漏或凭据使用。...详细分析 ▾
✓ 用途与能力
名称和描述（跨笔记本/VM的 SOTA 竞赛规划）与提供的脚本和参考匹配：工具创建竞赛、候选者、排行榜、运行卡、VM 启动清单，并渲染摘要。声明的二进制要求（python/python3）合适。
ℹ 指令范围
SKILL.md 指示运行本地 Python 脚本来创建和管理竞赛工件，并明确推荐将输出保持在竞赛工作空间内，并为实际运行使用配对的执行路径。范围限制为规划/记录。注意：运行时依赖于 sanitize_* 帮助程序（sota_public_safety）来编辑/清洗路径和 URL — 审查该文件以确认清洗行为如预期。
✓ 安装机制
无安装规格或外部下载；该技能是指令 + 从技能包运行的本地 Python 脚本。清单中无网络获取或存档提取。
✓ 凭证需求
该技能声明无需环境变量或凭据，甚至在程序模板中编码 OAuth 首选策略。脚本读取和清洗文件路径，但不需要令牌或密钥。
✓ 持久化与权限
always:false 和用户可调用；该技能不请求持久的高权限或修改其他技能。它将工件写入用户指定的路径（竞赛根目录），如其目的所期望。
安全有层次，运行前请审查代码。
运行时依赖
无特殊依赖
版本
latestv1.4.12026/3/19
润饰公开技能文档的措辞。
● 无害
安装命令 点击复制
官方npx clawhub@latest install sota-agent
镜像加速npx clawhub@latest install sota-agent --registry https://cn.clawhub-mirror.com
技能文档
目标
将模糊的「击败基准」请求转化为结构化的竞赛：- 固定的目标指标和分割
- 明确的文献和排行榜快照
- 有界的再现计划
- 明确的浏览器、笔记本或 VM 执行路径
- 当笔记本或浏览器状态重要时的 GUI 证据
- 每次回答一个问题的消解
- 仅在主张经受审查后进行推广
...（中间内容省略，仅保留部分示例）...
Goal
Turn a vague "beat the benchmark" request into a disciplined campaign:
- fixed target metric and split
- explicit literature and leaderboard snapshot
- bounded reproduction plan
- explicit browser, notebook, or VM execution lane
- GUI evidence when notebook or browser state matters
- ablations that answer one question at a time
- promotion only when the claim survives review
This skill is the frontier-planning and candidate-selection layer.
For browser evidence, VM execution, and promotion artifacts, pair it with
data-science-cv-repro-lab instead of letting the campaign drift into ad hoc runs.
Use This Skill When
- the user wants a CV or DS system pushed toward state-of-the-art results
- the task involves reproducing or surpassing recent papers
- the workflow needs paper triage, leaderboard tracking, or claim review
- the workflow includes OpenClaw, Colab, Kaggle, browser-only notebook actions, or GUI-heavy pages
- the user needs experiment management across browser research, notebooks, local runs, and long GPU jobs
- the user wants GPU VM or notebook watchdog logic, artifact pulls, or browser evidence for a SOTA candidate
- the question is whether a candidate is a real SOTA step or only noise, leakage, or benchmark overfitting
If the campaign includes serious execution or release review, use this skill to choose and rank candidates,
then use data-science-cv-repro-lab as the execution lane.
Quick Start
- Freeze the claim target before touching recipes.
- Initialize the campaign records immediately.
python3 {baseDir}/scripts/init_sota_campaign.py --root --campaign-id --title </code>.
- Use <code>python3 {baseDir}/scripts/init_sota_leaderboard_snapshot.py --out <json> --task <task> --dataset <dataset> --metric <metric> --split <split></code>.
- Use <code>python3 {baseDir}/scripts/init_sota_paper_triage.py --out <json> --campaign-id <id> --task <task></code>.
- Use <code>python3 {baseDir}/scripts/init_sota_program.py --out <json> --campaign-id <id> --task <task> --dataset <dataset> --metric <metric> --split <split></code> when you need one machine-readable benchmark, rerun, delegation, and auth plan.
- Use <code>python3 {baseDir}/scripts/init_sota_candidate_card.py --out <json> --candidate-id <id> --campaign-id <id> --objective <goal></code>.
- If execution review depends on synced QA runs, runtime sweeps, or benchmark panels, store the paired <code>data-science-cv-repro-lab</code> review dashboard path in the program and candidate records before the claim review starts.
- If the execution path depends on a real browser or notebook UI, use <code>python3 {baseDir}/scripts/init_sota_browser_run_card.py --out <json> --target-url <url></code>.
- If the browser or notebook surface needs manual or visual QA, use <code>python3 {baseDir}/scripts/init_sota_validation_scorecard.py --out <json> --scorecard-id <id> --surface <surface></code>.
- If a Colab, Kaggle, or notebook export bundle matters, use <code>python3 {baseDir}/scripts/init_sota_artifact_manifest.py --out <json> --bundle-root <dir></code>.
- If a long GPU VM run is involved, use <code>python3 {baseDir}/scripts/init_sota_vm_bootstrap_manifest.py --out <json> --output-root <run_root> --model-family <name> --command python train.py --epochs 40</code>.</p><ul><li>Separate the campaign roles even if one agent performs all of them.</li>
</ul>- Scout: papers, leaderboards, repos, and benchmark rules.
- Reproducer: baseline and top-paper reproduction.
- Ablator: controlled change sets and compute allocation.
- Reviewer: contamination, metric drift, and claim integrity.
- Promoter: final claim or hold decision.
- Keep the benchmark definition and final claim wording fixed.
- Use bounded scouting and review lanes for literature triage, repo inspection, per-paper extraction, and hard-case review.
- For repeated audits, batch over a manifest or CSV instead of free-form context accumulation.</p><ul><li>Pick the execution lane explicitly.</li>
</ul>- Browser or GUI lane: OpenClaw, Colab, Kaggle, or another real browser session when notebook UI state matters.
- Colab or notebook GPU lane: runtime selection, smoke validation, artifact export, and browser evidence.
- GPU VM lane: long runs with heartbeats, watchdogs, sync, and auto-stop policy.
- Local lane: cheap falsification, tiny reruns, and artifact review.</p><ul><li>Keep file writes inside one campaign workspace.</li>
</ul>- Create one dedicated campaign root and keep every <code>--out</code>, <code>--bundle-root</code>, and <code>--output-root</code> path under it.
- Do not point the bundled scripts at unrelated home-directory or system paths.
- Treat <code>scripts/sota_public_safety.py</code> as the canonical public-redaction layer for URLs, refs, and paths.</p><ul><li>Work the SOTA ladder in order.</li>
</ul>- Freeze the benchmark definition and auth rule before using more compute.
- Reproduce the trusted baseline first.
- Reproduce one relevant reference result or a close public checkpoint.
- Build a hypothesis backlog from literature gaps, not vibes.
- Run narrow ablations before broad recipe churn.
- Stress the best candidate on the fixed review surfaces.</p><ul><li>Claim only on full-surface wins.</li>
</ul>- Fixed benchmark score
- Reproduced baseline delta
- Compute or cost context
- Browser or GUI evidence if that lane mattered
- Failure-case review
- Exact evidence bundle
- Render the final review with <code>python3 {baseDir}/scripts/render_sota_claim_summary.py --candidate-card <json> --out <md></code>.</p><h2>Operating Rules</h2><h3>Campaign rules</h3><ul><li>One campaign has one target benchmark contract.</li>
<li>Do not let the target metric or split drift midstream.</li>
<li>Keep a short hypothesis backlog and kill low-information ideas quickly.</li>
<li>Record why each experiment exists before running it.</li></ul><h3>Codex multi-agent rules</h3><ul><li>Main thread owns the benchmark contract, stop conditions, and final claim decision.</li>
<li>Subagents should do bounded work only: scout, reproduce, ablate, or review.</li>
<li>Do not let one exploratory thread silently rewrite the campaign contract.</li>
<li>For repeated claim checks or literature extraction, prefer manifest-driven fanout over conversational drift.</li></ul><h3>Literature rules</h3><ul><li>Read only the papers or repos that change the candidate plan.</li>
<li>Extract the minimum useful fields: task, metric, split, data, compute, architecture, augmentations, training tricks, and caveats.</li>
<li>Prefer a reproduced strong baseline over copying five tricks from five papers without control.</li>
<li>Do not treat leaderboard rows as ground truth without checking task definition and split rules.</li></ul><h3>Ablation rules</h3><ul><li>Change one meaningful variable at a time when the goal is causal understanding.</li>
<li>If several knobs move together, label the run as a package change, not an ablation.</li>
<li>Keep one canonical baseline recipe alive for comparison.</li>
<li>Require the first winning candidate to survive at least one rerun or adjacent-seed check before escalating the claim.</li></ul><h3>Compute rules</h3><ul><li>Spend cheap compute on reproduction and short falsification first.</li>
<li>Do not push a long run unless the hypothesis would matter if it wins.</li>
<li>Record training cost, wall time, and hardware for every serious candidate.</li>
<li>Cut branches that cannot plausibly clear the target with the remaining budget.</li></ul><h3>OAuth and auth rules</h3><ul><li>Use ChatGPT or Codex OAuth-backed sessions as the default and preferred path.</li>
<li>Prefer Codex multi-agent or app-server workflows over orchestrators that require paid API keys.</li>
<li>Do not require or recommend <code>OPENAI_API_KEY</code>, other vendor API keys, or paid inference APIs as the default campaign runtime path.</li>
<li>If a third-party framework only works through paid API keys, treat it as reference material unless it can run fully through local tools and OAuth-backed Codex sessions.</li></ul><h3>OpenClaw browser rules</h3><ul><li>Use OpenClaw for public papers, leaderboards, docs, notebook-only steps, and GUI-heavy flows when the browser lane adds evidence.</li>
<li>Prefer direct public URLs over uploads or private sessions.</li>
<li>Capture leaderboard, notebook, or GUI evidence as notes, screenshots, and exact URLs when they are part of the claim path.</li>
<li>Fail hard on dead browser attach, missing notebook readiness, or unavailable requested model or runtime mode.</li>
<li>Treat screenshots and GUI evidence as supporting artifacts, not the claim itself.</li>
<li>Do not use browser-only summaries as the claim itself; claims still require benchmark artifacts.</li></ul><h3>Colab and notebook GPU rules</h3><ul><li>Select the accelerator explicitly before running expensive cells.</li>
<li>Run a smoke cell that proves imports, runtime, data mounts, and export paths all work.</li>
<li>Keep one stable export root and pull the artifact manifest plus at least one preview back locally.</li>
<li>Add the browser run card and validation scorecard when the notebook GUI is part of the evaluation story.</li></ul><h3>GPU VM rules</h3><ul><li>Create a named run root before launch.</li>
<li>Write a machine-readable VM bootstrap manifest before long runs.</li>
<li>Run long jobs under a heartbeat, session, or supervisor so liveness is explicit.</li>
<li>Sync metrics, summaries, and checkpoints back to a trusted store on a schedule.</li>
<li>Do not promote directly from live VM state; promote from synced artifacts and review evidence.</li></ul><h3>Claim safety rules</h3><ul><li>No SOTA claim without a fixed metric, split, and baseline.</li>
<li>No SOTA claim on a contaminated benchmark or hidden train-on-test path.</li>
<li>If the execution story depends on a dashboard or synced review surface, keep the dashboard path, source audit, and leakage audit in the claim packet.</li>
<li>If a candidate wins only on one slice while regressing important surfaces, hold it.</li>
<li>Report uncertainty honestly: "best internal result so far" is not the same as "new SOTA".</li>
<li>Small deltas need rerun or adjacent-seed support before they become claim language.</li></ul><h2>References</h2><p>Read only the reference that matches the task:</p><ul><li><code>references/sota-campaign-playbook.md</code></li>
</ul>- Full campaign structure, role separation, and stop conditions.
<ul><li><code>references/sota-program-rules.md</code></li>
</ul>- Rules for queues, stage discipline, ablations, and promotion gating.
<ul><li><code>references/campaign-harness-and-oauth-stack.md</code></li>
</ul>- What to reuse from Codex subagents, harness engineering, OpenEvolve, Symphony, Paperclip, and OptiLLM under an OAuth-only campaign rule.
<ul><li><code>references/benchmark-discipline.md</code></li>
</ul>- How to avoid contamination, metric drift, and invalid comparisons.
<ul><li><code>references/paper-triage.md</code></li>
</ul>- How to filter papers and extract only decision-relevant details.
<ul><li><code>references/openclaw-research-lane.md</code></li>
</ul>- How to use OpenClaw productively for public literature and leaderboard work.
<ul><li><code>references/openclaw-browser-lane.md</code></li>
</ul>- How to run GUI-heavy notebook, browser, and screenshot-based execution safely.
<ul><li><code>references/colab-vm-operations.md</code></li>
</ul>- How to manage Colab, Kaggle, and GPU VM execution lanes with smoke tests and artifact discipline.
<ul><li><code>references/claim-safety.md</code></li>
</ul>- Review rules for whether a candidate deserves a SOTA claim at all.
<ul><li><code>references/public-safety.md</code></li>
</ul>- Publication review rules for secrets, private refs, and raw notebook paths.</p><h2>Bundled Scripts</h2><ul><li><code>scripts/sota_public_safety.py</code></li>
</ul>- Pure local helpers for path, URL, ref, env, and command redaction. No network I/O or subprocess execution.
<ul><li><code>scripts/init_sota_campaign.py</code></li>
</ul>- Create a reusable campaign folder with benchmark, program, agent, research, leaderboard, plan, ablation, evidence, and claim files.
<ul><li><code>scripts/init_sota_program.py</code></li>
</ul>- Create a machine-readable program record with the fixed benchmark, baselines, rerun policy, bounded subagent roles, and OAuth rules.
<ul><li><code>scripts/init_sota_leaderboard_snapshot.py</code></li>
</ul>- Create a machine-readable snapshot of the target benchmark contract and current reference scores.
<ul><li><code>scripts/init_sota_paper_triage.py</code></li>
</ul>- Create a machine-readable literature queue for paper screening and extraction.
<ul><li><code>scripts/init_sota_browser_run_card.py</code></li>
</ul>- Create a sanitized browser evidence record for OpenClaw, Colab, Kaggle, or other notebook UI runs.
<ul><li><code>scripts/init_sota_validation_scorecard.py</code></li>
</ul>- Create a machine-readable GUI or notebook validation scorecard when visible state matters to the campaign.
<ul><li><code>scripts/init_sota_artifact_manifest.py</code></li>
</ul>- Create a machine-readable export-bundle manifest for notebook or VM artifact pulls with redacted public path metadata.
<ul><li><code>scripts/init_sota_candidate_card.py</code></li>
</ul>- Create a machine-readable card for a serious candidate, its execution lane, auth mode, and claim state.
<ul><li><code>scripts/init_sota_candidate.py</code></li>
</ul>- Create a machine-readable candidate record with change set, risks, and redacted public artifact refs.
<ul><li><code>scripts/init_sota_ablation_queue.py</code></li>
</ul>- Create a focused ablation queue for one candidate family.
<ul><li><code>scripts/init_sota_vm_bootstrap_manifest.py</code></li>
</ul>- Create a machine-readable bootstrap manifest for long GPU VM or cluster runs with public-release redaction.
<ul><li><code>scripts/update_sota_scoreboard.py</code></li>
</ul>- Refresh a ranked scoreboard for a fixed metric and goal direction.
<ul><li><code>scripts/init_sota_review_packet.py</code></li>
</ul>- Join the core artifacts for a promotion, hold, or cut decision.
<ul><li><code>scripts/render_sota_claim_summary.py</code></li>
</ul>- Render a concise markdown review from the machine-readable candidate card.
<ul><li><code>scripts/render_sota_program_summary.py</code></li>
</ul>- Render a concise markdown summary from the program, candidate, scoreboard, and review packet.
</p></div></div>
<div style="text-align:center;padding:var(--spacing-4);font-size:12px;color:var(--color-gray-400)">数据来源：<a href="https://clawhub.ai/zack-dev-cm/sota-agent" target="_blank" style="color:var(--color-gray-400)">ClawHub ↗</a> · 中文优化：龙虾技能库</div>
</section>
<section class="cta-card" aria-label="定制服务">
<div style="display:flex;align-items:center;gap:16px;flex-wrap:wrap">
<div style="flex:1;min-width:280px">
<div style="font-size:var(--font-size-md);font-weight:700;color:var(--color-text);margin-bottom:6px"><svg width="1em" height="1em" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" style="vertical-align:-0.125em"><path d="M14.7 6.3a1 1 0 0 0 0 1.4l1.6 1.6a1 1 0 0 0 1.4 0l3.77-3.77a6 6 0 0 1-7.94 7.94l-6.91 6.91a2.12 2.12 0 0 1-3-3l6.91-6.91a6 6 0 0 1 7.94-7.94l-3.76 3.76z"/></svg> OpenClaw 技能定制 / 插件定制 / 私有工作流定制</div>
<p style="font-size:var(--font-size-sm);color:var(--color-warning);margin-bottom:0;line-height:1.5"><svg width="1em" height="1em" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" style="vertical-align:-0.125em"><path d="M10.29 3.86 1.82 18a2 2 0 0 0 1.71 3h16.94a2 2 0 0 0 1.71-3L13.71 3.86a2 2 0 0 0-3.42 0z"/><line x1="12" y1="9" x2="12" y2="13"/><line x1="12" y1="17" x2="12.01" y2="17"/></svg> 免费技能或插件可能存在安全风险，如需更匹配、更安全的方案，建议联系付费定制</p>
</div>
<a href="/custom" class="btn-cta">了解定制服务</a>
</div>
</section>
</article>
<script>
function switchLang(lang){
document.querySelectorAll('.i18n').forEach(function(el){
var v=el.getAttribute('data-'+lang);
if(v)el.innerHTML=v;
});
document.querySelectorAll('.lang-cn').forEach(function(el){el.style.display=lang==='cn'?'block':'none'});
document.querySelectorAll('.lang-en').forEach(function(el){el.style.display=lang==='en'?'block':'none'});
document.querySelectorAll('.sk-lgb').forEach(function(b){b.classList.remove('sk-lga')});
var btn=document.getElementById('lang-'+lang);if(btn)btn.classList.add('sk-lga');
}
</script>
<style>
.sk-hero{border:1px solid var(--color-border);border-radius:12px;padding:22px;background:var(--color-bg-card);margin-bottom:16px;box-shadow:0 4px 12px rgba(0,0,0,.08)}
.sk-hero-main{display:flex;gap:24px}.sk-hero-left{flex:1;min-width:0}.sk-hero-right{width:200px;flex-shrink:0;display:flex;flex-direction:column;gap:10px}
.sk-tr{display:flex;align-items:center;gap:10px;flex-wrap:wrap;margin-bottom:4px}.sk-t{font-size:20px;font-weight:700;margin:0;line-height:1.3}
.sk-d{font-size:14px;color:var(--color-text-secondary);line-height:1.6;margin:8px 0 12px}
.sk-sr{display:flex;align-items:center;gap:8px;flex-wrap:wrap;font-size:13px;color:var(--color-text-secondary)}.sk-dot{color:var(--color-gray-400);opacity:.5}
.sk-au{font-size:13px;color:var(--color-text-muted);margin-top:4px}.sk-au a{color:var(--color-primary)}
.sk-lbox{font-size:13px;color:var(--color-text-secondary)}.sk-lbl2{font-size:11px;font-weight:600;color:var(--color-text-muted);margin-bottom:2px}
.sk-bdl{display:block;text-align:center;min-height:48px;line-height:20px;border-radius:12px;font-size:1.05rem;font-weight:600;padding:0.75rem 2rem;transition:background 0.2s}
.sk-bdl:hover{text-decoration:none;color:#fff;background:#333}
.btn--dark.sk-bdl{background:#1a1a1a;color:#fff;border:none}
.btn--dark.sk-bdl:hover{background:#333;color:#fff;transform:translateY(-1px);box-shadow:0 4px 12px rgba(0,0,0,0.15)}
.btn--dark.sk-bdl:active{transform:translateY(0);box-shadow:none}
[data-theme="dark"] .btn--dark.sk-bdl{background:#e0e0e0;color:#1a1a1a}
[data-theme="dark"] .btn--dark.sk-bdl:hover{background:#ccc;color:#1a1a1a;box-shadow:0 4px 12px rgba(255,255,255,0.08)}
[data-theme="dark"] .btn--dark.sk-bdl:active{background:#bbb;box-shadow:none}
.sk-sc{border:1px solid var(--color-border);border-radius:12px;padding:16px;margin-bottom:16px;background:var(--color-bg-card)}
.sk-sc--safe{border-left:4px solid #67C23A}
.sk-sc--suspicious{border-left:4px solid #E6A23C}
.sk-sc--dangerous{border-left:4px solid #F56C6C}
.sk-sch{font-size:15px;font-weight:700;margin-bottom:12px}
.sk-scr{display:flex;align-items:center;gap:12px;padding:8px 12px;background:var(--color-gray-100);border-radius:8px;margin-bottom:8px}
.sk-scw{font-size:13px;font-weight:600;min-width:120px}.sk-scs{font-size:13px;font-weight:700}.sk-scl{font-size:12px;color:var(--color-primary);margin-left:auto}
.sk-scsm{font-size:13px;color:var(--color-text-secondary);line-height:1.6;padding:8px 12px;background:var(--color-gray-100);border-radius:8px;margin-bottom:8px}
.sk-scd{margin-top:4px}.sk-scd summary{cursor:pointer;color:var(--color-primary);font-size:13px;font-weight:600;padding:4px 0}
.sk-dm{padding:8px 12px;background:var(--color-gray-100);border-radius:8px;font-size:13px;margin-bottom:4px}
.sk-adv{margin-top:8px;padding:12px;background:#FDF6EC;border:1px solid #FFEEBA;border-radius:8px}.sk-adv ol{margin:4px 0 0 16px;font-size:12px;color:var(--color-text-secondary);line-height:1.8}
.sk-ig{display:grid;grid-template-columns:repeat(auto-fit,minmax(240px,1fr));gap:12px;margin-bottom:16px}
.sk-ib{border:1px solid var(--color-border);border-radius:12px;padding:16px;background:var(--color-bg-card)}.sk-ib h3{font-size:14px;font-weight:700;margin:0 0 12px}
.sk-lic-desc{font-size:12px;color:var(--color-text-secondary);line-height:1.5;margin:8px 0 0}
.sk-ii{font-size:13px;margin-bottom:4px;display:flex;align-items:center;gap:8px}.sk-il{color:var(--color-text-muted);min-width:50px;flex-shrink:0;font-size:12px}
.sk-ii code{background:var(--color-gray-100);padding:1px 6px;border-radius:4px;font-size:12px}
.sk-cm{font-family:monospace;font-size:13px;padding:12px;background:var(--color-gray-100);border:1px solid var(--color-border);border-radius:8px;margin-bottom:8px;cursor:pointer;word-break:break-all;transition:border-color 300ms}
.sk-cm:hover{border-color:var(--color-primary)}.sk-cl{font-size:11px;font-weight:600;color:var(--color-text-muted);display:block;margin-bottom:2px}
.sk-cma{background:var(--color-primary-light);border-color:var(--color-primary);color:var(--color-primary)}
.sk-cnn{background:#FDF6EC;border:1px solid #FFEEBA;border-radius:12px;padding:16px;margin-bottom:16px}.sk-cnn h3{font-size:15px;margin:0 0 8px}.sk-cnn p{font-size:14px;line-height:1.6;margin:0}
.sk-lgb{padding:4px 14px;border-radius:9999px;border:1px solid var(--color-border);background:var(--color-bg-card);color:var(--color-text-secondary);cursor:pointer;font-size:13px;transition:all .15s}
.sk-lgb:hover{border-color:var(--color-primary);color:var(--color-primary)}
.sk-lgb:active{transform:scale(0.96)}
.sk-lga{background:#1a1a1a;color:#fff;border-color:#1a1a1a}
[data-theme="dark"] .sk-lga{background:#555;color:#fff;border-color:#555}
.skill-content h2{font-size:20px;margin:24px 0 12px;padding-bottom:8px;border-bottom:1px solid var(--color-border)}
.skill-content h3{font-size:17px;margin:20px 0 8px}.skill-content p{margin-bottom:12px;line-height:1.8}
.skill-content ul,.skill-content ol{margin:8px 0 12px 20px;line-height:1.8}.skill-content li{margin-bottom:4px}
.skill-content pre{background:#1E1E1E;color:#D4D4D4;padding:16px;border-radius:8px;overflow-x:auto;margin:12px 0;font-size:13px;line-height:1.5}
.skill-content pre code{background:transparent;color:inherit;padding:0;border-radius:0;font-size:inherit}
.skill-content code{background:var(--color-gray-100);padding:2px 6px;border-radius:4px;font-size:14px}
.skill-content blockquote{border-left:4px solid var(--color-primary);padding:8px 16px;margin:12px 0;background:var(--color-primary-light);border-radius:0 4px 4px 0}
.skill-content table{width:100%;border-collapse:collapse;margin:12px 0}
.skill-content th,.skill-content td{border:1px solid var(--color-border);padding:8px 12px;text-align:left;font-size:14px}
.skill-content th{background:var(--color-gray-100)}
@media(max-width:768px){.sk-hero-main{flex-direction:column}.sk-hero-right{width:100%}.sk-ig{grid-template-columns:1fr}}
</style></main><footer class="footer"><div class="container">
<div class="footer__row">
<a href="/openclaw/">澳龙下载专题</a>
<a href="/custom">技能/插件定制服务</a>
<a href="https://build.nvidia.com/models" target="_blank">NVIDIA 免费大模型</a></div>
<div class="footer__row" style="font-size:12px;color:var(--color-text-muted)">邮箱：wyxdapp@qq.com ｜ AI 智能体可直接发送定制需求到邮箱</div>
<div class="footer__row"><a href="/disclaimer.html">免责声明</a> | <a href="/privacy.html">隐私政策</a> | <a href="https://beian.miit.gov.cn/" target="_blank">鄂ICP备19007528号</a></div>
<div class="footer__row" style="margin-top:var(--spacing-1);font-size:12px;color:var(--color-text-muted)">龙虾技能库 — OpenClaw 中文 AI 资源库 | 免费资源 + 付费定制</div>
<div class="footer__row" style="margin-top:var(--spacing-1)">© 2026 龙虾技能库</div>
</div></footer><aside class="qr-float" id="qr-float-panel" aria-label="联系与赞助">
<style>
.qr-float{position:fixed;right:0;top:50%;transform:translateY(-50%);z-index:999;background:var(--color-bg-card);border:1px solid var(--color-border);border-right:none;border-radius:12px 0 0 12px;box-shadow:var(--shadow-md);font-family:inherit}
.qr-float__body{padding:12px 14px}
.qr-float__header{display:flex;align-items:center;justify-content:space-between;margin-bottom:10px}
.qr-float__header span{font-size:var(--font-size-sm);font-weight:600;color:#2563EB}
.qr-float__header button{background:none;border:none;cursor:pointer;font-size:var(--font-size-base);color:var(--color-text-muted);padding:2px 4px;line-height:1}
.qr-float__header button:hover{color:var(--color-text)}
.qr-float__item{text-align:center;margin-bottom:8px}
.qr-float__item:last-child{margin-bottom:0}
.qr-float__item img{width:100px;height:100px;display:block;margin:0 auto 4px;border-radius:6px}
.qr-float__item span{font-size:var(--font-size-xs);color:var(--color-text-muted)}
.qr-float__trigger{display:none;position:fixed;right:0;top:50%;transform:translateY(-50%);z-index:999;align-items:center;justify-content:center;width:40px;height:40px;background:var(--color-bg-card);border:1px solid var(--color-border);border-right:none;border-radius:12px 0 0 12px;box-shadow:var(--shadow-md);cursor:pointer;color:#2563EB;font-size:20px}
.qr-float__trigger:hover{background:var(--color-bg-hover,var(--color-bg-card))}
@media(max-width:768px){
.qr-float__body{display:none}
.qr-float__trigger{display:flex}
.qr-float__item img{width:80px;height:80px}
}
</style>
<div class="qr-float__body" id="qr-float-body">
<div class="qr-float__header">
<span>联系 & 赞助</span>
<button onclick="document.getElementById('qr-float-body').style.display='none';document.getElementById('qr-float-trigger').style.display='flex'" aria-label="收起面板">✕</button>
</div>
<div class="qr-float__item"><img src="/image/erweima.png" alt="站长微信二维码" loading="lazy"><span>关注站长微信</span></div>
<div class="qr-float__item"><img src="/image/weixinpay.png" alt="微信赞助收款码" loading="lazy"><span>微信赞助</span></div>
<div class="qr-float__item"><img src="/image/alipay.png" alt="支付宝赞助收款码" loading="lazy"><span>支付宝赞助</span></div>
</div>
<button class="qr-float__trigger" id="qr-float-trigger" onclick="document.getElementById('qr-float-body').style.display='block';document.getElementById('qr-float-trigger').style.display='none'" aria-label="展开联系与赞助面板"><svg width="1em" height="1em" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" style="vertical-align:-0.125em"><path d="M21 15a2 2 0 0 1-2 2H7l-4 4V5a2 2 0 0 1 2-2h14a2 2 0 0 1 2 2z"/></svg></button>
</aside></body></html>