Data & Analytics

白盒加密分析：白盒实现识别、密钥提取。 触发词：白盒、whitebox、白盒加密、密钥隐藏

MIPS 架构逆向与路由器固件分析方法论：延迟槽、$gp 调用约定、大小端判断、httpd 定位。 触发词：MIPS、mipsel、路由器固件、嵌入式、延迟槽、big-endian、大端、小端、httpd、busybox

Ghidra 工作流：导入→自动分析→反编译→脚本化。 触发词：Ghidra、反编译、headless

覆盖率引导模糊测试：AFL++/libFuzzer/honggfuzz、插桩与语料初始化、 afl-fuzz 运行参数、覆盖率(afl-cov)、字典/结构化输入。 触发词：fuzz、模糊测试、AFL、afl-fuzz、libFuzzer、honggfuzz、 覆盖率、corpus、语料、dictionary、挖洞、fuzzing。

函数式语言运行时逆向（Haskell/OCaml）：闭包/堆对象模型、调用约定、数据流优先策略。 触发词：Haskell逆向、OCaml逆向、GHC RTS、thunk、STG、OCaml runtime、闭包、函数式产物。

Electron 桌面应用逆向：asar 解包、主/渲染进程 JS、V8 字节码（.jsc）边界、CDP 动态调试、反调试对抗。 触发词：Electron、asar、桌面应用逆向、.jsc、V8 快照、CDP、ELECTRON_RUN_AS_NODE、devtools 检测。

磁盘/文件系统取证：删除恢复、时间线、可疑文件定位。 触发词：磁盘取证、文件系统、删除恢复、时间线、ext4、NTFS

加密数据还原：定位解密函数、写解密脚本。 触发词：解密、decrypt、还原数据、解密流量

汽车逆向：CAN 总线、ECU 固件、AUTOSAR Classic（RTE/runnable/BSW）与 Adaptive（执行管理/清单）。 触发词：汽车、CAN、ECU、车载、OBD、UDS、AUTOSAR、RTE、runnable、SWC、BSW、E_OS_ACCESS、 Adaptive、ARA、Execution Management、Function Group、Machine State、Execution Manifest

AI 模型分析第一入口（分流器）：识别输入形态——模型文件走文件层逆向（re-ai-model）、 仅有 API 走行为层评估（re-ai-attack）、恶意行为走恶意分析。 触发词：AI模型分析、AI模型安全、模型文件、模型泄露、模型逆向、模型分析入口。

AI 模型文件逆向与静态分析：ONNX/PyTorch/Safetensors/TFLite 格式解析、 网络结构还原、权重提取、文件级水印分析（权重 pattern/metadata/tensor hash/embedding 异常）。 触发词：模型文件、权重提取、ONNX解析、safetensors、pth分析、pt文件、模型结构还原、文件级水印

AI 模型安全评估与取证（行为层）：extraction assessment（API 黑盒提取评估）、 fingerprint verification（行为指纹与归属验证）、privacy leakage evaluation（成员推断/隐私泄露评估）、 robustness evaluation（对抗鲁棒性评估）。 触发词：API模型窃取、黑盒模型复制、模型指纹、成员推断、对抗样本、行为水印、模型安全评估、隐私泄露评估。

Build a live Cowork artifact (persistent HTML widget) backed by a Coupler.io dataflow. Use this skill whenever the user wants a live dashboard, persistent widget, daily-check page, or interactive explorer over Coupler.io data — including phrases like "live artifact for Coupler", "Coupler dashboard widget", "build a widget over a Coupler dataset", "Coupler.io live dashboard", "build a daily dashboard from a Coupler dataflow", or any time they ask to render Coupler data in a re-openable view th...

Use this skill when the user wants to analyze sales pipeline, review win rates, investigate sales velocity or cycle length, build a sales report, evaluate rep performance, look at lead source quality, or answer questions about their CRM data. Triggers include 'how is the pipeline', 'win rate by segment', 'sales cycle', 'pipeline coverage', 'rep performance', 'forecast accuracy', 'why are deals stalling', 'lead source quality', 'stage conversion', 'pull my Salesforce numbers', 'HubSpot deals r...

Use for "where am I wasting money on TikTok", "which ad groups should I pause", "what should I turn off", "which campaigns should I scale", "clean up my TikTok account", "find me negative keywords for Search Ads", or "what do I cut and where does the money go" — even without the words "waste" or "scale". Also use when someone wants a budget-neutral reallocation proposal for the account. TikTok Ads only.

Use for "is my TikTok account built right", "why are my ad groups stuck in learning", "do I have too many ad groups", "should I consolidate", "why won't this ad group exit learning", "I inherited this TikTok account, what's wrong with it", or "my delivery keeps restarting" — even when the user never says "structure". Covers how the account is organised and what that costs it in learning. TikTok Ads only.

Use for "is Pangle wasting my money", "should I turn off automatic placement", "how is Search Feed doing", "which countries are working on TikTok", "is iOS or Android converting", "what time of day should I be running", or "where is my budget actually going" — even when the user never says "placement". Covers where and when the ads are shown, and what each of those is worth. TikTok Ads only.

Use for "can I trust my TikTok conversion numbers", "why does TikTok claim more purchases than Shopify", "are my conversions double counted", "is my pixel firing", "why do standard and real-time conversions disagree", "my conversions dropped overnight", or "how much spend has no tracking" — and whenever someone doubts the numbers, even without the word "audit". Run before trusting any cost or return conclusion about the account. TikTok Ads only.

Use for "how are my TikTok ads doing", "why did my cost per result go up", "what changed on TikTok this month", "which campaigns improved", "my CPMs are climbing", "my video views dropped", or a weekly or monthly account check — even when the user never says "review". Also use when someone wants the baseline read before deciding anything else in the account. TikTok Ads only.

Use for "how is GMV Max doing", "should I use GMV Max or run campaigns manually", "is GMV Max cannibalising my other campaigns", "which products are selling on TikTok Shop", "my GMV Max ROI dropped", "I can't see anything inside GMV Max", or "is TikTok Shop worth the spend" — even when the user only says "Shop ads" or "my TikTok black box". TikTok Ads only, and only for accounts running TikTok Shop.

Use for "is my TikTok creative burning out", "my hook rate is dropping", "my CPMs keep rising", "how long before I need new videos", "which ads should I refresh", "how much new creative do I need a month", "how many videos should we shoot", or "this video used to work" — even when the user never says "fatigue". This is the how-long-has-it-left read and the production-cadence number; for which videos win in the first place, use the creative analysis skill. TikTok Ads only.

Use for "which creative is working on TikTok", "which videos should I put money behind", "what should the next round of creative look like", "is my hook working", "which video is holding attention", "why do people scroll past my ad", or "what do my winners have in common" — even when the user never says "creative". This is the which-videos-win read; for which winning videos are dying, use the fatigue skill instead. TikTok Ads only.

Use when someone asks for a monthly TikTok report for a client, wants the TikTok section of a client deck built, needs last month's paid social written up, asks for an end-of-month TikTok report, asks "what do I tell the client", or asks how to present a target they missed — even when they never say the word "report". For agencies and in-house teams reporting upward. TikTok Ads only.

Use for "am I on track with my TikTok budget", "will I overspend this month", "how much should I be spending a day", "which ad groups are capped", "we underspent and I don't know why", "why won't this ad group spend", or a mid-month spend check — even when the user never says "pacing". Also use when someone needs to know whether adding budget to an ad group would do anything at all. TikTok Ads only.