AI & ML

jonathaneciana17-maker Skill

Caching strategies for LLM prompts including Anthropic prompt caching, response caching, and CAG (Cache Augmented Generation)

format-technologies Bundle

Find your best case study candidates and draft case studies from customer conversation data — before you ever schedule an interview. Use when the user wants to write a case study, customer story, or success story. Also triggers on "case study," "customer story," "who should we feature," "which customer has a good story," "case study draft," "case study candidate," "social proof," "write up a customer win," "we need case studies," "I don't know where to start with case studies," or "turn customer data into a case study." This skill uses customer intelligence data (calls, emails, support tickets, CRM) to identify which customers have the strongest stories and then produces a working draft based on what they've already said — so you walk into the interview with a near-finished piece instead of a blank page. Works best with a customer intelligence platform like Format (MCP server) but also works manually with compiled transcripts and recordings.

format-technologies Bundle

Use when a product manager or engineer wants to ground a ticket (Linear, Jira, or any tracker) in real customer evidence using Format MCP. Canonical invocation: 'using the Format MCP and the format-ticket-research skill, research this ticket' followed by pasted ticket text or a ticket reference. Also triggers on 'what are customers saying about this ticket', 'who's asking for this', 'find customer evidence for [ticket]', 'what requirements are customers implying for [feature]'. Produces a compact evidence page: customer asks grouped by the distinct need expressed, verbatim quotes with links to every piece of evidence, and what the evidence suggests the ticket should account for. It presents evidence and context rather than grading demand or recommending whether to build.

format-technologies Bundle

Use when a product manager or engineer wants to ground a ticket (Linear, Jira, or any tracker) in real customer evidence using Format MCP. Canonical invocation: 'using the Format MCP and the cube-ticket-research skill, research this ticket' followed by pasted ticket text or a ticket reference. Also triggers on 'what are customers saying about this ticket', 'who's asking for this', 'find customer evidence for [ticket]', 'what requirements are customers implying for [feature]'. Produces a compact evidence page: customer asks grouped by the distinct need expressed, verbatim quotes with links to every piece of evidence, and what the evidence suggests the ticket should account for. It presents evidence and context rather than grading demand or recommending whether to build.

format-technologies Bundle

Use when you are about to write a report into Format with create_report or replace_report — the craft guide for authoring one well. Read it BEFORE your first authoring write, not after. Covers the block-tree document model and which block does which job, the entity-mention law (companies, people and records are linked chips, never plain text), evidence discipline for quoting real customers, the cover fields (title, scope, conclusion, tldr), the reach-bar chart convention, the image upload handshake, and the draft → publish → send lifecycle. Triggers on 'write this up as a report', 'put this in Format as a report', 'create a report from these findings', and on any revision of a report you authored.

format-technologies Bundle

Use when assessing the selling competency of one sales rep from their real call recordings in Format. Triggers include: 'coach [rep]', 'score [rep]'s calls', 'how is [rep] selling', 'competency assessment for our AEs', 'sales coaching', 'review call quality', 'where is [rep] losing deals', or 'rep scorecard'. The skill requires a single named rep and a time window — it stops and asks if either is missing — then reads every real conversation that rep had in the window via the Format MCP, classifies each by reachability and motion, scores the genuine prospecting conversations against a behavioural rubric, and produces a competency assessment with verbatim evidence and one prioritised coaching focus. Assessment only — it surfaces evidence and a single priority; it does not write call scripts or run the coaching session.

format-technologies Bundle

Use when assessing the selling competency of one sales rep from their real call recordings in Format. Triggers include: 'coach [rep]', 'score [rep]'s calls', 'how is [rep] selling', 'competency assessment for our AEs', 'sales coaching', 'review call quality', 'where is [rep] losing deals', or 'rep scorecard'. The skill requires a single named rep and a time window — it stops and asks if either is missing — then reads every real conversation that rep had in the window via the Format MCP, classifies each by reachability and motion, scores the genuine prospecting conversations against a behavioural rubric, and produces a competency assessment with verbatim evidence and one prioritised coaching focus. Assessment only — it surfaces evidence and a single priority; it does not write call scripts or run the coaching session.

jimmyvgdy Bundle

服务端应用、API、业务逻辑、认证鉴权、应用事务、并发、任务、Worker、资源生命周期或后端代码审查时使用，覆盖 Java、Python、Node.js、Go、.NET、Rust 及其他后端技术栈。先识别实际运行时和框架；纯数据库/中间件/基础设施、纯浏览器前端或纯模型/RAG 语义任务不要使用。

jimmyvgdy Bundle

任务跨会话、多阶段、多模块、多仓库、多 Agent、生产观察期、上下文可能压缩，或需持续维护目标、进度、证据、决策和交接时使用。简单一次性任务不要触发，也不要为无状态工具调用重复写检查点。

jimmyvgdy Bundle

代码、脚本、迁移或逻辑配置发生行为变化，或任务涉及测试、Git、CHANGELOG、审批、部署、重启、灰度、回滚和生产操作时使用。只读解释、无行为变化文字修改和普通文档排版通常不触发；多 Agent 复审仅按风险组合。

jimmyvgdy Bundle

Use for server-side applications, APIs, business logic, authentication and authorization, application transactions, concurrency, jobs, workers, resource lifecycle, or backend reviews across Java, Python, Node.js, Go, .NET, Rust, and other stacks. Identify the actual runtime first. Do not use for pure data/infrastructure, browser-only, or pure model/RAG semantics work.

jimmyvgdy Bundle

高风险实施前设计审查，或行为改动后的独立复审、回归兼容、安全、性能、数据契约、并发状态和测试证据审查时使用。简单低风险、证据充分或无行为变化任务不要触发，也不要为了形式固定多开 Reviewer。

jimmyvgdy Bundle

Use for cross-session, multi-stage, multi-module, multi-repository, multi-agent, observation-window, and context-compaction work that needs recoverable checkpoints.

jimmyvgdy Bundle

Use for risk-based design gates and independent functional, compatibility, security, performance, contract, concurrency, and delivery reviews.

jimmyvgdy Bundle

Use only for cross-task retrospectives, observation governance, model-cost routing, reviewer calibration, routing drift, proposal lifecycle, project isolation, and assistant version governance.

thatmike1 Bundle

Read another agent session's transcript straight from its log on disk, with zero footprint on the observed session, and list which sessions are running right now. Covers Claude Code, Codex and the Antigravity CLI (agy/Gemini), and the T3 Code threads wrapped around any of them. Use before sending a message to another session, when observing or coaching a session that must stay untouched, when asked which T3 threads are still open or unsettled, or when asked to peek at, read, or follow what another session is doing.

thatmike1 Bundle

Answer questions about the user's other running Claude Code sessions — what they are, which one to touch next, what a confusing one is saying, what to send it, and starting a new one. Use when the user asks what their sessions are doing, where to start, which session had X, to explain or unstick another session, to write the next prompt for one, or to put a session to work on something. Also load on "jarvis".

thatmike1 Bundle

Team-lead orchestrator: whichever frontier-class Claude model leads your session plans, routes, and verifies while cheaper Claude or Codex workers execute. Use for: orchestrate, delegate, conductor mode, save tokens, multi-agent.

thatmike1 Bundle

Write a copy-paste prompt that hands one or more tasks (usually beads issues) to a fresh Claude instance to run autonomously while the user is away — no questions, no waiting for input. Includes picking which tasks are safe to run unattended. Use when the user says they're going AFK (shower, lunch, errand, overnight), asks "which tasks can just run without me", or wants an autonomous/batch prompt. For attended sessions the user will steer, use the live-prompt skill instead.

thatmike1 Bundle

Write a copy-paste prompt that hands a task (usually a beads issue) to a fresh Claude instance the user will actively work with, in either collaborative (discuss-approach-first) or off-the-leash (implement-end-to-end) mode. Use when the user says "give me a prompt", "prompt for <task-id>", or wants to start a fresh instance on a task. For unattended runs (user going AFK), use the afk-prompt skill instead.

tsetsugekka Bundle

Reusable public-safe sentiment and market-emotion framework for A-shares, Japanese stocks, US stocks, indexes, and sector themes. Use when a stock or market move needs emotion-cycle classification, main-line versus follower judgment, expectation-gap analysis, forum/news sentiment synthesis, risk-on/risk-off context, market-sector-stock resonance, post-surge continuation analysis, A-share sector/theme constituent mapping such as 哪些股票/相关股/概念股/龙头股, or when other stock skills need a shared sentiment layer. Supports optional user-specified private RAG folders without storing or publishing private materials.

yananlong Bundle

Improve the flow, shape, and readability of existing prose while preserving substance, specificity, tone, and constraints, including planned multi-agent section, paragraph-range, or chunk-level revision for substantial documents when workers are available. Use when Codex needs to make explanatory writing less choppy, less list-heavy, less fragmented, or less over-paragraphed; combine related sentences or paragraphs aggressively but readably; add connective reasoning; favor conjunctions, relative clauses, subordinate clauses, and clean participial phrasing over semicolons; or turn terse notes into natural connected prose. Do not use when the content is inherently list-shaped, procedural, tabular, or governed by a stricter task-specific format.

yananlong Bundle

Revise existing technical writing into clear, audience-facing prose while preserving evidence, technical meaning, uncertainty, and scope, including planned multi-agent section or issue-cluster revision for substantial artifacts when workers are available. Use for academic manuscripts, industrial technical reports, whitepapers, evaluation summaries, methods/results/limitations prose, technical claims, contribution or decision framing, and terminology conversion from internal shorthand to reader-facing language. Use after critique from research-paper-review or research-review-loop when the task is to revise the artifact's prose and claims, not to re-review it. Do not use when the main task is generic prose flow, citation work, paper planning, novelty review, experiment design, market strategy, or writing unsupported claims.

yananlong Bundle

source-grounded market, competitor, procurement, funding, regulatory, and patent-landscape diligence for academic research commercialization. use when codex needs to search public market or patent information for a research asset, technology, prototype, dataset, method, device, or commercialization thesis; map incumbents, substitutes, buyers, assignees, patent families, citations, ownership signals, public funding, procurement demand, regulatory/reimbursement constraints, and evidence gaps; prepare diligence artifacts for commercialize-academic-research or related research skills.

managedcode Bundle

Primary router skill for broad .NET work. Classify the repo by app model and cross-cutting concern first, then switch to the narrowest matching .NET skill instead of staying at a generic layer.

managedcode Bundle

Use the open-source free `CSharpier` formatter for C# and XML. Use when a .NET repo intentionally wants one opinionated formatter instead of a highly configurable `dotnet format`-driven style model.

magnus-gille Bundle

Close a coding-agent session or produce its final handoff only when the user explicitly asks to close, wrap up, or produce an end-of-session handoff.

magnus-gille Skill

Adversarial Debate (cross-model)

magnus-gille Skill

Delegate a bounded one-shot task to Hugin's orchestrator-v1 broker — cheaper / smaller runtimes via the `hugin-mcp` MCP. Use for summarize / extract / classify / draft / rewrite / short reason work where you'd rather not spend Claude tokens. Submits, polls to completion, surfaces the result, and prompts the user to rate it.

magnus-gille Skill

Delegate a self-contained task to Magnus's M5 when local inference is requested or the project's delegation policy applies.

magnus-gille Bundle

Resolve a batch of GitHub tickets via parallel headless Claude Code sessions — one isolated git worktree per ticket, gated permissions, triple review (Codex + local-model first-pass + orchestrator comment), interrupt-resume, and merge orchestration. Use when the user wants many well-scoped tickets solved autonomously across one or more repos.

magnus-gille Bundle

Run a two-round adversarial review from Codex with Claude as the independent reviewer for a consequential decision or substantive draft when cross-model critique is warranted.

magnus-gille Bundle

Run a requested, read-only cross-model review of a pull request or branch diff using the installed Codex CLI. Use only when the user asks for review-pr-codex or a Codex PR review.

vamseeachanta Skill

Model Selection Skill

jeandoom Skill

连接和管理 Teambition（项目管理系统），通过 Teambition MCP 服务实现任务、项目、成员等数据查询和管理。当用户提到 Teambition、项目管理、任务管理等时使用此 skill。

jeandoom Skill

本地项目异常分析（code agent 模式）。分析和解决当前目录所在项目的 bug：直接用本地代码分析，不拉码、不建目录、不产物料文件，分析结果和解决方案全部直接对话输出。由 bug-killer 模式判断后调度，也支持独立触发。触发词：本地 bug、当前项目报错、这个项目异常、native 分析。