Use when the user asks to "build the norm card for this platform", "what are the char limits and visible-fold cutoffs here", "is the LinkedIn link-in-first-comment thing documented or folklore", or "which of our platform cards are stale"; maintains the dated, versioned per-platform norm cards in the references/platforms/ pack — char limits, visible-fold cutoffs, hashtag norms, format/aspect specs, link and first-comment placement, disclosure-label mechanics, algorithm emphases (e.g. 小红书 search+saves weighting) — every row labeled platform-documented (official doc, Measured) or Estimated-folklore (named source) with last-verified and review-by dates, and any card past its review date flagged stale rather than trusted. Extends the single pack in place; never forks a second one. Not for picking which channels to run — use channel-portfolio-planner. 平台规范卡/字符限制/折叠线/话题标签/算法侧重/过期标记

Use when the user asks to "plan the participation ramp before we promote", "how much account history or karma do we need in this community", or "design entry incentives and member lifecycle for our own Discord"; produces the per-community pre-promotion warming plan — account-history/tenure expectations (Estimated, named sources), a give-before-ask ledger spec, a per-community etiquette + rule digest with last-verified dates, and the warming → active graduation criteria that channel-registry requires as state-transition evidence — plus the owned-community variant (entry paths + member lifecycle for your own Discord/Slack/forum/企业微信私域). Not for launch-day submissions or T-0 threads — use community-launch-runner. 社区预热/先给后取/账号养成/毕业标准

Use when the user asks to "script this short video", "write a TikTok / Reels / Shorts script", "给这条抖音或视频号视频写脚本", or "fix the hook — viewers drop off in the first seconds"; produces timestamped beat-sheet scripts on the retention-gate model (0-2s hook / 2-5s confirmation / 5-15s payoff / loop-or-CTA) with per-scene script lines, on-screen text for muted viewing, asset keywords, and 9:16 format-param rows per platform norm card for TikTok, Reels, Shorts, 抖音, and 视频号 — plus 2-3 hook options from named hook families and an AI-content disclosure line by default on realistic synthetic media. Spec-only: rendering, TTS, and publishing stay with the user's own tools — no pipelines, no upload automation. Not for creator video briefs — use brief-generator; long-form video SEO belongs to content-writer. 短视频脚本/抖音分镜/开头钩子/竖屏9:16

Use when the user asks to "design an employee advocacy program", "set up founder-led sharing", or "build a share kit for the team"; produces an advocacy program blueprint in two modes — participation-driven opt-in (default) or top-down assigned with its coercion and authenticity risks flagged — with a voluntary opt-in roster spec submitted as channel-registry proposal events, share kits with mandatory per-person variation, staggered human posting windows plus anti-pod guardrails (no coordinated identical reshares, no engagement rings), per-person material-connection disclosure lines per FTC and 《互联网广告管理办法》, and a Slack/Teams distribution spec. Not for paid creator campaigns — use campaign-planner. 员工倡导/创始人IP分享/内部分享计划/披露合规

Use when the user asks to "track rankings" or "查排名"; measures keyword and SERP-position deltas over time from provided exports or connected tools, including AI-response checks. Not for multi-metric stakeholder reports or setting alerts — use performance-monitor (report or alert mode). 排名追踪/SERP监控

Use when the user asks to "log this subscriber's opt-in", record unsubscribes/complaints, or query lawful basis; curates pseudonymous consent facts through the append-only consent stream and applies suppression/erasure tombstones immediately. Not for SEND scoring — use email-quality-auditor; not for building segments — use list-segment-builder. 订阅同意台账/实时退订抑制/合法性依据登记

Use when the user asks to "map what our surfaces say today", "inventory our current messaging", or "find the gap between what we say and what we mean"; produces the narrative baseline — a surface-by-surface inventory of what every owned touchpoint (homepage, pricing, docs, decks, social bios, email footers) claims RIGHT NOW, each line labeled Measured / User-provided / Estimated, plus a per-surface gap read vs the intended message and the drift-baseline snapshot the Evaluate phase measures future drift against. Not for authoring the canon — use message-system-architect; not for scoring the surfaces or running the vetoes — use narrative-quality-auditor. 现状叙事盘点/各触点口径/意图差距/漂移基线

Use when the user asks to "codify our brand voice", "define naming rules for our products and tiers", or "write the tone-of-voice guide with banned phrases"; produces the brand-level voice canon (register, tone spectrum, banned-phrase list, few-shot examples drawn only from the brand's own published material) and the naming tax (product / feature / tier naming rules plus approved and banned terms) that seeds the narrative-registry canon and that every channel's voice adaptation points up to. Not for per-platform voice adaptation — use channel-registry's voice-dossier; not for finished copy or blog posts — use content-writer; not for the message hierarchy itself — use message-system-architect; not for claim adjudication — use offer-claims-registry. 品牌语气/词汇表/命名税/禁用词/品牌语言规范

Use when the user asks to "keep the launch momentum going after launch week", "plan a changelog / release-notes cadence as GTM", or "is this update worth a relaunch"; produces a T+1→T+30 momentum plan — a launch-moment calendar (milestone / shipped-loop / badge moments only), announcement-tier routing (major = full-channel, medium = targeted, minor = changelog-only), a relaunch legitimacy call, spike-to-owned handoff briefs, and the next Tier-1 moment with launch-stacking spacing. Not for the 30-day content-reuse map or paid amplification execution — use content-amplifier; not for planning the next launch end to end — use launch-tier-planner. 抗第二周断崖/changelog-as-GTM/relaunch/下一发布时刻

Use when the user asks to "run a launch retro / post-mortem", "compare launch results vs targets by channel", or "decide what to keep or kill for the next launch"; produces a structured D1/W1/M1 retrospective — a per-channel actual-vs-target table (UTM-attributed own analytics as the truth column, platform self-reported numbers as reference, every figure labeled Measured / User-provided / Estimated), a 5-Whys chain on the single largest miss, keep / kill / change decisions per channel, 3-5 actionable learnings for the next launch, and an outcome snapshot submitted to the launch registry. Not for return math (CPA / ROI) — use roi-calculator; not for the stakeholder-facing report writeup — use report-generator; not for a metric deep-dive — use performance-analyzer. 发布复盘/渠道归因/5-Whys/keep-kill

Use when the user asks to "triage launch feedback", "cluster reviews, comments, and board posts into themes", or "set up a you asked, we shipped loop"; produces a feedback theme digest (frequency, severity, representative quotes per theme), an open→planned→started→completed/declined status loop with duplicate-merge and notification rules, shipped-change announcement material, and a compliant social-proof harvest protocol (never incentivized store reviews). Not for repurposing or amplifying the harvested proof — use content-amplifier; not for executing testimonial outreach threads — use outreach-manager. 反馈分诊/状态环/社证收割/评测合规

Use when the user asks to "run my launch day", "build a launch day runbook / war room", or "decide CONTINUE or ROLLBACK after the push"; produces a pre-conditions gate check (launch-readiness-auditor SHIP verdict + the authoritative date in launch-registry — missing either stops the skill), a dated hour-blocked runbook with owners (morning irreversible pushes, daytime monitoring loop, evening consolidation), a forced observation-window verdict after every irreversible action against pre-declared kill criteria, a P0-P3 incident ladder with rollback playbooks, and T-0 status lines for the registry proposal protocol. Not for channel submission content and platform rules — use community-launch-runner; not for media replies — use press-media-relations. 发布日runbook/作战室/观察窗/回滚裁决/发布日指挥

Use when the user asks to "plan an influencer campaign", "build a campaign blueprint", "track or close a creator campaign", or "record a late campaign correction"; produces the plan and, when requested, a non-canonical evidence tracker with scoped identity, publication, reconciliation, close, and reopen receipts. Not for individual creator briefs — use brief-generator; not for overall product launches without creators — use launch-tier-planner; not for sending, publishing, amplifying, or paying — use the owning execution workflow. 达人营销策划/种草方案/活动追踪与关账

Use when the user asks to "create an influencer brief" or "write a campaign brief"; produces a structured creator brief with deliverables, key messages, creative direction, timeline, disclosure rules, and compensation terms. Not for choosing how to split spend across creators — use budget-optimizer. 达人合作简报/创作者BF

Use when the user asks to "build email segments from my list", "make engaged / lapsed / RFM segments", "set up cart-abandoner or lifecycle-stage audiences", or "build a suppression list of unsubscribes and bounces"; turns the user's OWN list/CRM/GA4/ecommerce export into behavioral, attribute, and lifecycle-stage segments plus a suppression list, with per-segment sizes labeled Measured/Estimated, informing the SEND E (Engagement/targeting) dimension. Not for scoring EQS or running vetoes — use email-quality-auditor; not for authentication or spam-content checks — use deliverability-qa. 邮件列表分群/生命周期分群/抑制名单/流失召回

Use when the user asks to "pre-launch check the landing page", "run a Quality-Score preflight", or "verify ad-to-page message match before launch"; produces an ad↔page continuity report — message-match gaps, above-the-fold check, page-speed read, form-friction count, mobile-render flags — as a pass/fix punch list. Not for redesigning or rewriting the page — use landing-optimizer; not for scoring the account or the RQS — use ad-account-auditor. 落地页体验预检/广告落地页一致性检查

Create or update feature specifications from natural language descriptions. Use when starting new features or refining requirements. Generates spec.md with user stories, functional requirements, and acceptance criteria following spec-driven development methodology.

Provides guidance for training and analyzing Sparse Autoencoders (SAEs) using SAELens to decompose neural network activations into interpretable features. Use when discovering interpretable features, analyzing superposition, or studying monosemantic representations in language models.

World-class data science skill for statistical modeling, experimentation, causal inference, and advanced analytics. Expertise in Python (NumPy, Pandas, Scikit-learn), R, SQL, statistical methods, A/B testing, time series, and business intelligence. Includes experiment design, feature engineering, model evaluation, and stakeholder communication. Use when designing experiments, building predictive models, performing causal analysis, or driving data-driven decisions.

This skill should be used when working with reinforcement learning tasks including high-performance RL training, custom environment development, vectorized parallel simulation, multi-agent systems, or integration with existing RL environments (Gymnasium, PettingZoo, Atari, Procgen, etc.). Use this skill for implementing PPO training, creating PufferEnv environments, optimizing RL performance, or developing policies with CNNs/LSTMs.