method-candidate-ab-validity-trigger-confound
A/B qualification candidate distilled verbatim by Qwen3.8-27B from the working-set receipts teacher episode; experimental candidate only, not promoted.
Skill by deyi2026
Chalk Labs
Need help deploying method-candidate-ab-validity-trigger-confound?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/deyi2026-method-candidate-ab-validity-trigger-confound
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/method-candidate-ab-validity-trigger-confound
What's inside
- Skill name
method-candidate-ab-validity-trigger-confound- Description
- A/B qualification candidate distilled verbatim by Qwen3.8-27B from the working-set receipts teacher episode; experimental candidate only, not promoted.
- License
- Apache-2.0
- Instructions
- 36 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Qwen-authored candidate (verbatim) FRICTION - 首个 OFF 因环境缺 providers.json 走 unknown provider，属 invalid arm，不能参与 A/B 结论。 - ON 的 fold 到第 15 个大 read_file 才真实触发，前段与 OFF 行为近似，机制未早触发。 - elapsed/cache-hit 的 ON 是第二个 run，同物理模型且无安全 cache clear，存在 warm/order confound。 DISCRIMINATOR - 有效 arm 判据：同 model/prompt SHA/tools/reasoning settings + 独立 session/data dir + 环境可解析 provider。 - 机制触发判据：必须观测到 fold 实际发生（folded_results 非空/达到阈值），而非仅配置开启。 - 收益判据：任务正确性（4/4、15/15）与收敛轮次/token 分开看；elapsed/cache 在 warm/order 未控时只作 observation。 SHORTEST_PATH - 先做 arm 有效性预检（provider 可解析、配置 SHA 一致、独立目录），失败即标 invalid 不进入对比。 - 在 ON 侧埋点确认机制真实触发点（第几次 read 触发 fold、folded 数量），未触发则该 run 对机制无效。 - 对 elapsed/cache 类指标，要么做 cache clear/顺序平衡，要么降级为 observation，不用于 promotion 决策。 GENERAL_RULE - A/B 结论前，先排除 invalid arm（环境/配置不可解析）与未触发机制的 run，再谈收益。 - 机制收益需“触发证据 + 任务正确性 + 收敛/成本”三者齐备；任一缺失则 mixed/不足 promotion。 - 时序相关指标（elapsed、cache-hit）在 warm/order 未受控时不得作为因果或 promotion 依据。 COUNTEREXAMPLE - 若把首个 unknown-provider 的 OFF 当作有效基线，会高估 ON 的相对收益。 - 若只看 ON 配置开启就断言 fold 生效，会忽略前 14 次 read 未触发的空窗。 - 若把第二个 run 的 elapsed/cache 优势归因于 ON，会混入 warm/order confound。 METHOD CARD - 名称：working-set receipts A/B 蒸馏 - 适用：验证某机制（如 fold）是否带来真实收益的 A/B - 步骤：1) arm 预检（provider/SHA/独立目录）2) 埋点确认机制触发 3) 分维度判收益 - 判据：机制 PASS=触发证据；任务 PASS=正确性+可恢复性；收益=收敛/成本且无 confound - 降级：elapsed/cache 未控 warm/order 时仅 observation - 反例：invalid arm 入对比、未触发即断言、warm run 归因 - 产出：PASS/mixed/不足 promotion 三态结论，附触发点与 confound 说明
From deyi2026/llm-first-loop, Apache-2.0.
More from deyi2026/llm-first-loop
- 0.589score
integrity-auditor
deyi2026
论文诚信审计——对一篇论文（外部 PDF/DOI/arXiv 或本地写作产出）做图像误用、数值异常、逻辑缺口三轨证据审计，产出 Level 1–4 分级、可复核的 audit_report.md。用于投稿前自查、审稿前评估、对 AI 生成内容做"数字有没有来源、图表有没有被动过"级核验。触发：integrity…
SkillAFreshness: A★ 1Claude Code +9 - 0.589score
fit-page-limit
deyi2026
篇幅达标规划——论文超页限或篇幅不足时，逐节量化"长度去哪了"，给出优先级方案：压缩/移附录哪些低价值节、补实哪些薄弱节，同时保护贡献、结果与引用；不删证据、不为凑页注水。触发：超页限、页数超了、压缩论文、该删哪里、trim、over the page limit、too long、too short、fit…
SkillAFreshness: A★ 1Claude Code +9 - 0.589score
anonymize-paper
deyi2026
双盲匿名深扫 + camera-ready 干净还原——投稿前扫尽泄漏面（作者块/grant 致谢/PDF 元数据/第一人称自引/仓库与主页链接/LaTeX 注释/.bib 注释/主目录路径/补充材料），自带 stdlib…
SkillAFreshness: A★ 1Claude Code +9 - 0.588score
gzh-design
deyi2026
微信公众号文章排版引擎，将 Markdown 转换为可直接粘贴到公众号编辑器的 HTML。主题风格从 references/theme-index.md 注册的自定义主题库中选取，自动章节编号、关键词下划线标记、引言卡片、目录导航、代码块、图片/GIF、作者签名。支持 Markdown / Word(.docx)…
SkillAFreshness: A★ 1Claude Code +9 - 0.584score
assess-paper
deyi2026
论文整体健康度评估——"我这论文现在什么水平/能不能投/先改哪里/优缺点"时使用。编排本仓库已装的 preflight-check、verify-citations、anonymize-paper、fit-page-limit 做桌拒合规/引用/匿名/篇幅检查，汇总成一份先优点、再风险、带优先级修改清单的…
SkillAFreshness: A★ 1Claude Code +9 - 0.584score
method-ab-experiment
deyi2026
设计或审查模型、Agent、缓存、上下文、工具策略、runtime 优化的 A/B/qualification 实验时使用。用于锁唯一自变量、隔离 cache/session/evidence/model residency 等跨 run…
SkillAFreshness: A★ 1Claude Code +9 - 0.584score
method-candidate-current-callsite-refresh
deyi2026
Repo/API candidate distilled verbatim by Qwen3.8-27B from the parallel-line stale-callsite teacher episode; experimental candidate only, not promoted.
SkillAFreshness: A★ 1Claude Code +9 - 0.584score
method-candidate-hydrate-trace
deyi2026
Qwen3.8-27B 从离线网页绕路 episode 经 teacher exemplar 蒸馏出的实验 candidate；仅用于独立 transfer qualification，不代表已 promotion。适用于当前页面是 loading shell 且显式引用可能负责加载目标内容的脚本时，测试是否应先沿…
SkillAFreshness: A★ 1Claude Code +9 - 0.584score
method-candidate-relative-clock-never-state
deyi2026
Root Cause candidate distilled verbatim by Qwen3.8-27B from the fresh-process relative-clock sentinel teacher episode; experimental candidate only, not…
SkillAFreshness: A★ 1Claude Code +9 - 0.584score
code-nav
deyi2026
使用仓库已有的 Pyright language server 做 Python 跨文件语义导航。用于改函数前确认调用者/被调用者、收敛 grep 结果到真实语义引用、检查符号爆炸半径，或在陌生模块里追踪引用链。必须通过 readiness gate 才接受语义结果；把 LSP semantic scope 与…
SkillAFreshness: A★ 1Claude Code +9 - 0.581score
merge-review-gate
deyi2026
合并 main 审查闸（RULE-AI-34）——分支 Goal/任务收口前的专门合并审查子代理任务书：核完成证据、核分支卫生、核 main 基线、批准并入 main、落盘回执
SkillAFreshness: A★ 1Claude Code +9 - 0.580score
humanizer-zh-next
deyi2026
编辑或审阅已有中文文本中的空话、重复和模板化表达，保留原意、事实、确定程度和作者声音。 主文件包含通用表达模式、中文句式、体裁判断、作者样本使用和文件保护，适用于去 AI 腔与自然化润色。 商务营销、论文与审稿回复、基金申请分别按需加载专用 reference。用于写作质量编辑，不判定作者身份。
SkillAFreshness: A★ 1Claude Code +9
Signals
- Stars
- 1
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Sep 8, 2026 (1 month ago)
- Last release
- Oct 22, 2026 (8 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- Apache-2.0
- Repository
- github.com/deyi2026/llm-first-loop
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from deyi2026
- 0.589score
anonymize-paper
deyi2026
双盲匿名深扫 + camera-ready 干净还原——投稿前扫尽泄漏面（作者块/grant 致谢/PDF 元数据/第一人称自引/仓库与主页链接/LaTeX 注释/.bib 注释/主目录路径/补充材料），自带 stdlib…
SkillAFreshness: A★ 1Claude Code +9 - 0.589score
integrity-auditor
deyi2026
论文诚信审计——对一篇论文（外部 PDF/DOI/arXiv 或本地写作产出）做图像误用、数值异常、逻辑缺口三轨证据审计，产出 Level 1–4 分级、可复核的 audit_report.md。用于投稿前自查、审稿前评估、对 AI 生成内容做"数字有没有来源、图表有没有被动过"级核验。触发：integrity…
SkillAFreshness: A★ 1Claude Code +9 - 0.589score
fit-page-limit
deyi2026
篇幅达标规划——论文超页限或篇幅不足时，逐节量化"长度去哪了"，给出优先级方案：压缩/移附录哪些低价值节、补实哪些薄弱节，同时保护贡献、结果与引用；不删证据、不为凑页注水。触发：超页限、页数超了、压缩论文、该删哪里、trim、over the page limit、too long、too short、fit…
SkillAFreshness: A★ 1Claude Code +9 - 0.588score
gzh-design
deyi2026
微信公众号文章排版引擎，将 Markdown 转换为可直接粘贴到公众号编辑器的 HTML。主题风格从 references/theme-index.md 注册的自定义主题库中选取，自动章节编号、关键词下划线标记、引言卡片、目录导航、代码块、图片/GIF、作者签名。支持 Markdown / Word(.docx)…
SkillAFreshness: A★ 1Claude Code +9 - 0.584score
method-ab-experiment
deyi2026
设计或审查模型、Agent、缓存、上下文、工具策略、runtime 优化的 A/B/qualification 实验时使用。用于锁唯一自变量、隔离 cache/session/evidence/model residency 等跨 run…
SkillAFreshness: A★ 1Claude Code +9 - 0.584score
assess-paper
deyi2026
论文整体健康度评估——"我这论文现在什么水平/能不能投/先改哪里/优缺点"时使用。编排本仓库已装的 preflight-check、verify-citations、anonymize-paper、fit-page-limit 做桌拒合规/引用/匿名/篇幅检查，汇总成一份先优点、再风险、带优先级修改清单的…
SkillAFreshness: A★ 1Claude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get method-candidate-ab-validity-trigger-confound running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
method-candidate-ab-validity-trigger-confound skill. Agent Mart. https://agentmart.dev/skills/deyi2026-method-candidate-ab-validity-trigger-confound (last updated 2026-10-24)