AI & ML

richlander Bundle

Experiment harness shell for the custom MCP delivery environment on the Markout task. Carries no inline grounding; curated package context is delivered only via the get_package_context MCP tool when the agent chooses to call it (gate under test = GROUNDING_GATE).

richlander Bundle

Experiment harness shell for the MCP delivery environment (multi-package triage). Carries no inline grounding; package context for any referenced package is delivered only via the get_package_context MCP tool when the agent chooses to call it.

richlander Bundle

The complete Markout coding-agent integration guide: installation, the required 3-part source-generation pattern (annotate, register on a partial MarkoutSerializerContext, serialize through it), the full attribute reference (including MarkoutLink, MarkoutValueMap, MarkoutSection.GroupBy, MarkoutMaxItems, MarkoutBoolFormat), every built-in shape type (Metric, Breakdown, Callout, TreeNode, Description, CodeSection), and formatter selection (Markdown, TSV, plain text, Spectre). Markout has NO reflection fallback, so a MarkoutSerializerContext is always required.

richlander Bundle

Experiment harness shell for the MCP delivery environment (resident-API probe). Carries no inline grounding; System.Text.Json context is delivered only via the get_package_context MCP tool when the agent chooses to call it.

richlander Skill

Use when System.Text.Json work is not a plain POCO round-trip — inspecting/mutating JSON without a model (including answering a few questions about an unknown payload), editing JSON whose shape is not controlled by the app, reading or writing at high throughput (Utf8JsonReader / Utf8JsonWriter, byte and Stream overloads), or processing a JSON array/feed too large to fit in memory (DeserializeAsyncEnumerable).

richlander Bundle

Experiment harness shell for the MCP delivery environment. Carries no inline grounding; package context is delivered only via the get_package_context MCP tool when the agent chooses to call it.

richlander Skill

Use when System.Text.Json runs under Native AOT or trimming (PublishAot / PublishTrimmed), or when you want the faster, reflection-free serialization path — i.e. a JsonSerializerContext with [JsonSerializable]. Reflection-based JsonSerializer compiles but THROWS at run time under AOT; the source generator is the only supported path. Also covers configuring a context with [JsonSourceGenerationOptions] and why naming policy is per-context, not per-model, so one type serialized by two contexts can emit two different spellings, plus word-valued enums that must remain trim/AOT-safe.

richlander Skill

Use when targeting recent .NET runtimes (8/9/10+) and you need the newer, stricter System.Text.Json behavior or APIs a model trained on older docs may not know — the JsonSerializerOptions.Strict preset, AllowDuplicateProperties, RespectNullableAnnotations / RespectRequiredConstructorParameters, rejecting unknown/unmapped configuration properties, PipeReader overloads, and JsonMarshal. Also covers why duplicate-key rejection matters at untrusted parse boundaries, the .NET 11 [JsonNamingPolicy] attribute and its non-transitivity, and getting newer STJ APIs on an older runtime via the out-of-band package.

richlander Bundle

Experiment harness shell for the MCP delivery environment (ambiguous multi-package triage). Carries no inline grounding; package context for any referenced package is delivered only via the get_package_context / summarize_package_context MCP tools when the agent chooses to call them.

agentscope-ai Skill

Calculate PI/PID controller gains using Internal Model Control (IMC) tuning rules for first-order systems.

agentscope-ai Skill

Fit first-order dynamic models to experimental step response data and extract K (gain) and tau (time constant) parameters.

agentscope-ai Bundle

Build unified multi-level category taxonomy from hierarchical product category paths from any e-commerce companies using embedding-based recursive clustering with intelligent category naming via weighted word frequency analysis.

crozzbite Skill

Standards for proactive error auditing and issue correlation using Sentry MCP.

crozzbite Skill

OWASP LLM Top 10 security controls for AI/LLM applications. Trigger: When implementing LangChain flows, agent tools, or LLM outputs.

crozzbite Skill

Security hub skill orchestrating all security standards (OWASP ASVS, LLM Top 10, GDPR, PCI DSS). Trigger: When auditing security, handling secrets, sanitizing PII, or reviewing PRs.

kao273183 Bundle

AI / LLM 應用品質評估專屬流程。覆蓋幻覺（hallucination）/ 事實性（groundedness）、相關性、結構化輸出合法性、prompt injection 抵抗、安全/毒性、成本（$/req）、延遲（p95）、token 用量、回歸（eval set）、一致性、拒答校準、RAG 檢索品質。整合 promptfoo / DeepEval / Ragas / LLM-as-judge + golden dataset + deterministic seed。當使用者提到「LLM 測試 / AI 品質 / 幻覺 / hallucination / groundedness / prompt injection / eval / 評估集 / RAG 評估 / LLM-as-judge / 模型回歸 / AI app 品質 / token 成本」時觸發。配套：property-based-test-gen（fuzz prompt）、security-scan（injection 屬安全）、performance-test-gen（LLM API 延遲/壓力）、test-data-factory（eval 資料）、bug-report。

kao273183 Bundle

專為 Flutter 應用量身打造的自動化測試腳本生成 Skill。從 Google Sheet TC、JIRA 票號或功能描述生成可執行的 Dart 測試腳本，涵蓋 Unit Test（flutter_test）、Widget Test（WidgetTester）、Integration Test（integration_test）、Golden Test、Platform Channel Test。採用 Fake-over-Mock 原則、Robot Pattern / Page Object Model、Mocktail/Mockito 策略，並整合 Firebase Test Lab、`flutter test --coverage` 覆蓋率工具。當使用者提到「Flutter 自動化」、「寫 Dart 測試」、「把 Flutter TC 轉成程式碼」、「widget test 自動化」、「integration_test 腳本」、「Flutter UI test」、「patrol 測試」、「Golden test 自動化」，或針對 Flutter/Dart 專案要求生成自動化腳本時使用。與 test-automation 並存：test-automation 適用原生 iOS (XCUITest) / Android (Espresso)，flutter-test-automation 適用 Flutter/Dart 專案或 Flutter+Native 混合。

ryzhakar Bundle

Govern multi-agent research across any knowledge surface: technology ecosystems, market landscapes, academic fields, regulatory environments, curated indices, or any domain requiring breadth-first exploration followed by depth-first verification. Triggers: "research an ecosystem", "survey the landscape", "evaluate options for", "deep-dive", "compare alternatives", "map out what exists", "find the best X for Y", "audit the market", "what should I use for", "what's available in", "how does X compare to Y across the field".

ryzhakar Skill

Decompose work into agent-delegated units across model tiers. Agents are cheap, context is expensive — decompose aggressively, delegate everything, assemble results. Triggers: "delegate", "parallelize", "parallel launch", "launch", "orchestrate", "use agents for", "run in parallel"; or any task with independent subtasks.

catlog22 Bundle

Swarm intelligence team skill — ACO-driven multi-agent exploration with hybrid LLM coordinator + Python optimization controller. Coordinator generates swarm-config from user task, then runs K iterations of N parallel ants guided by pheromone state. Universal task space via config (nodes + scoring rule). Triggers on "team swarm", "swarm intelligence", "蚁群".

catlog22 Bundle

Universal skill diagnosis and optimization tool. Detect and fix skill execution issues including context explosion, long-tail forgetting, data flow disruption, and agent coordination failures. Supports Agy CLI for deep analysis. Triggers on "skill tuning", "tune skill", "skill diagnosis", "optimize skill", "skill debug".

catlog22 Bundle

Iterative skill tuning via execute-evaluate-improve feedback loop. Uses maestro delegate Claude to execute skill, Agy to evaluate quality, and Agent to apply improvements. Iterates until quality threshold or max iterations. Triggers on "skill iter tune", "iterative skill tuning", "tune skill".

catlog22 Skill

Quick execution for small tasks — minimal run lifecycle (start + done) with evidence recording. Full LLM capability, scoped to mechanically clear tasks. Arguments: <intent> [-y]

zxremail Skill

Extract durable working preferences from recent Cursor chats and convert them into skills, rules, subagents, or commands. Use when asked to learn preferences, mine feedback, personalize workflows, generate team/person-specific agent guidance, or when the user mentions 从聊天提炼、沉淀工作流、内化成 skill/agent/command、 回顾对话偏好、workflow-from-chats.

zxremail Skill

基于飞书群组和单聊的历史消息，自动生成指定成员的个人周工作报告。通过飞书 MCP 工具获取用户信息、群组列表和聊天记录，筛选该成员发送和被@的消息，分析整理为结构化周报。Use when the user asks to generate a personal weekly report (个人周报) from Feishu chat history, or mentions "基于飞书沟通内容生成周报".

madappgang Skill

Runs models through the claudish MCP tools — team, create_session, run_prompt — and resolves model IDs against the live catalog. Use when the user mentions claudish, OpenRouter, or external AI models.

madappgang Skill

Patterns and templates for building Claude Code plugins. Use for plugin development — creating a plugin, skill and agent templates, plugin architecture, or standardizing structure.

madappgang Skill

Multi-agent orchestration enforcement for /team command. Validates session directory paths and ensures /team Tasks use a valid agent from the routing whitelist. Use when debugging /team orchestration failures.

madappgang Skill

Track agent, skill, and model performance metrics for optimization. Use when measuring agent success rates, tracking model latency, analyzing routing effectiveness, or optimizing cost-per-task.

madappgang Skill

Reference for running models through the claudish MCP tools — team, create_session, run_prompt. Covers model routing, native Claude slots, require_pattern shape checks, and error handling. Use when working on /team, /delegate, or any claudish call.

madappgang Skill

Quick reference for running external models in orchestration. They are invoked via claudish MCP tools (team, create_session), never the CLI. Use when unsure how to specify an external model.

madappgang Skill

Runs the same task across multiple AI models in parallel and aggregates verdicts. Use when the user wants a second opinion, multi-expert validation, or consensus from Grok, Gemini, GPT-5, or Kimi.

madappgang Skill

Routes tasks to model tiers (haiku/sonnet/opus) by complexity. Use for model selection, complexity routing, picking a tier, or optimizing API cost.

madappgang Skill

MANDATORY tracking protocol for multi-model validation. Creates structured tracking tables BEFORE launching models, tracks progress during execution, and ensures complete results presentation. Use when running 2+ external AI models in parallel.

madappgang Skill

Prevents goal drift in long multi-agent runs by validating outputs against the original objective at checkpoints. Use when orchestrating 3+ agents or multi-phase work.

madappgang Skill

Coordinate multiple agents in parallel or sequential workflows. Use when running agents simultaneously, delegating to sub-agents, switching between specialized agents, or managing agent selection.