AI & ML

kesslernity Bundle

Sweeps a SharePoint library listing (an export or a listing the agent can reach, with file name, link, review-date column and owner column) for documents past their review date and returns a dated DRAFT review sweep report (Overdue, Due soon, No review date, Summary) plus one ready-to-paste reminder per document owner for the user to send. Never changes metadata, sends a message or invents a document, date or owner. Use when the user asks to "find the documents past their review date", "run the review sweep on the policies library", "which documents are overdue for review", "audit the review dates in this library export" or "draft reminders to owners about stale documents". Do not use for duplicate, conflict and ownership checks across knowledge files, use knowledge-base-hygiene-review instead; for gaps in an engineering deliverables register, use master-document-register-check. Drafts for human review; never approves, authorises or signs off.

kesslernity Bundle

Reviews the instructions text and skills of a declarative agent, with its stated capabilities, knowledge sources and audience, for prompt-injection exposure, data-leakage paths, over-broad permissions and missing refusals, and returns an attack-surface map, a findings table with severity, quoted evidence and attack path, one pasteable fix per finding, a probe pack the owner can run and a refusal-coverage table. Use when the user asks to "red-team this agent", "check these instructions for prompt injection", "could this agent leak data", "review the permissions on my agent" or "what should this agent refuse". Do not use for checking a SKILL.md against the format rules, use skill-file-reviewer instead; do not use for designing the launch test set, use agent-evaluation-plan instead. Drafts for human review; never approves, authorises or signs off.

kesslernity Bundle

Proposes the piping content a P&ID adds to an accepted PFD process model: line inventory with off-page continuity and required ancillary lines, line identification elements per the project line numbering procedure, connection items at nozzles matched to a quoted schedule, per quoted standards, specification break questions and observations. Every proposal cites its source and revision; no source, no proposal; never sizes a line or selects a class. Use when the user asks to "run the piping enrichment", "start the discipline enrichment", "list the lines the P&ID adds to this PFD" or "fill the line identifiers from the numbering procedure" after gate G2. Do not use for instruments, loops or instrument tags, use instrumentation-and-control-enrichment instead; do not use before gate G2, use process-model-check. Drafts for human review; never approves, authorises or signs off.

kesslernity Bundle

Checks an extracted PFD register as a process model and returns findings and unknowns only: connectivity closure, stream numbering, presence of design conditions, equipment duty text, phase and service labels, package and off-sheet scope, boundaries, operating cases, and agreement between the drawing and attached process documents, then holds at gate G2. Creates no design values. Use when the user asks to "check the process model", "verify connectivity" or "is the PFD ready for enrichment" after gate G1 and before enrichment. Do not use for extraction from the drawing, use pfd-intake-and-extraction instead; do not use to propose lines or instruments, use piping-enrichment. Drafts for human review; never approves, authorises or signs off.

takusaotome Bundle

AI活用提案の専門コンサルタントスキル。LLM・AIエージェントの活用事例知識を基に、業界別、部門別、シーン別に最適なAI導入提案を行います。システム化、業務改善、効率化を得意とし、具体的な活用事例、期待効果、導入ステップを含む実践的な提案を提供します。Use when proposing AI/LLM adoption strategies for specific industries, departments, or business scenarios.

desko77 Skill

1C:Naparnik (1С:Напарник) MCP server tools - BSL code analysis, review, rewriting, ITS knowledge base, platform documentation search, configuration documentation. Use when checking BSL code quality, searching ITS standards, reading platform docs, comparing versions, or finding configuration-specific documentation.

desko77 Bundle

EDT-based 1C:Enterprise (1С:Предприятие 8.3+) development via the EDT MCP server - BSL code analysis and editing, metadata inspection and construction, module navigation, query validation, managed forms, error checking, debugging, infobase update. Use when the project is an EDT workspace: 1C/BSL modules, .mdo metadata, 1C queries, managed forms. Not for 1C 7.7 (.1s/.ert/1Cv7.MD) and not for Configurator-format sources without an EDT project - those have their own skills.

desko77 Skill

Справочник по HTTP-API LM Studio (сервер локальных моделей, обычно порт 1234). Используй когда надо программно управлять моделями на LM Studio - узнать что загружено, загрузить или выгрузить модель, задать длину контекста и параллельность, отключить размышления модели, - а также при диагностике: модель thinking/reasoning не отключается, enable_thinking игнорируется, ответ приходит пустым а весь бюджет уходит в reasoning_content, запросы падают с HTTP 500 на холодном старте или при загрузке крупной модели, ошибка Context size has been exceeded, Model is unloaded, эндпоинт отвечает 404 или 200 с телом Unexpected endpoint, неверно определяется длина контекста. Содержит проверенные схемы эндпоинтов v1, параметр reasoning, особенности протокола и замеры на 30-35B.

desko77 Bundle

>

desko77 Bundle

Улучшает и структурирует короткие неструктурированные промпты и постановки задач, превращая их в подробные императивные ТЗ с пронумерованными шагами анализа, явными граничными случаями и описанием формата ожидаемого вывода. Сохраняет все термины и условия исходника, не добавляет новых требований. Используй когда пользователь просит: улучшить промпт, проработать постановку, расширить описание задачи, структурировать ТЗ, оформить заметку в подробную инструкцию, сделать промпт точнее или подробнее, подготовить ТЗ из черновика, причесать задание, преобразовать короткую заметку в понятное ТЗ. Принимает текст в аргументе команды или путь к .md файлу.

desko77 Bundle

Установка и обновление рабочего окружения агента: скилы, правила и команды этого набора плюс связанные инструменты (MCP-серверы, плагин EDT, локальная транскрибация, конвертеры документов, утилиты 1С). Работает и на чистой машине, и на уже настроенной - сначала снимает опись того, что стоит, потом показывает план и ставит ТОЛЬКО выбранное. Используй когда пользователь просит поставить или обновить скилы, настроить окружение с нуля, перенести набор на новую машину, доставить недостающие инструменты, проверить что из окружения отвалилось. Триггеры: установить скилы, настроить окружение, поставить на новой машине, обновить набор, доставить недостающее, что у меня стоит, проверить окружение, перенести конфигурацию агента.

desko77 Bundle

Применять при переписывании текстов, сгенерированных LLM-агентами (отчеты, README, доки, письма, посты), в живой человеческий стиль. Триггеры - пользователь пишет 'убери AI-стиль', 'перепиши по-человечески', 'сделай естественно', 'убери воду', 'не как ChatGPT', 'убери LLM-штампы'; либо на входе текст с явными маркерами генерации: H2/H3 на каждый абзац, буллеты вместо прозы, штампы 'в современном мире', 'давайте погрузимся', избыток длинных тире, эмодзи-заголовки, обязательные 'надеюсь, это поможет!'. Сверх того ВСЕГДА проверяет технический регистр: в README, журнале изменений, документации и описаниях заменяет бытовую метафору на действие и предмет ('досыпает', 'кладет', 'копит', 'под капотом', 'из коробки', 'ядро библиотеки'); в письме, ответе и эссе эта проверка сама выключается по жанру, ключ отключения - --no-technical. Переписывание в живой стиль не применять к структурированным форматам, где списки и заголовки уместны по существу: API-документация, чек-листы, табличные данные, бенчмарки, юридические док

desko77 Skill

Practical guide for creating human-readable and agent-parseable diagrams using Mermaid. Includes conservative, renderer-compatible templates and when-to-use guidance.

desko77 Bundle

Локальная транскрибация аудиофайлов без отправки в облако. Используй когда пользователь просит транскрибировать запись, расшифровать аудио, сделать конспект встречи, преобразовать речь в текст. Только для аудио (m4a/mp3/wav/ogg/flac/aac/wma/opus). Движок: faster-whisper CUDA + опц. диаризация sherpa-onnx GPU (CUDA, RTF ~0.24). Поддерживает разделение по спикерам.

jorinyang Bundle

Use when optimizing agent skills, evaluating skill quality, or running skill review — 技能自主优化器：9维评分体系 + 棘轮机制 + 人在回路。触发词：优化skill、skill评分、自动优化、skill质量检查、达尔文、darwin、帮我改改skill、skill怎么样、提升skill质量、skill review、skill打分。

jorinyang Skill

DingTalk Workspace CLI (dws) — 安装、配置、20项MCP服务全能力边界（ai表格/审批/日历/文档/通讯录/邮箱/考勤等）。适用于钉钉生态内的自动化操作。

jorinyang Skill

即梦/CapCut AI视频与图片生成 — CLI安装、token管理、文生视频/图生视频/omni_reference。内置情绪导演 Prompt 工程核心方法论（物理优先、六段式结构、情绪→生理转化），接入AI Native Workflow

jorinyang Skill

Delegate coding tasks to external AI coding agent CLIs: Claude Code, OpenAI Codex, and OpenCode. Provider comparison, mode selection, and orchestration patterns via Hermes terminal.

jorinyang Bundle

Firecrawl 自托管 web 搜索与抓取——搜索网页、抓取内容、爬取网站、提取结构化数据时自动使用。MCP 调用本地 Firecrawl (localhost:3002)，无限额度。触发：搜索/查资料/抓取/爬取/提取数据/web search/scrape/crawl/extract。

jorinyang Bundle

Agent Skill 三维评测引擎。自动或手动对 Skill 进行"执行精准度 × 端到端时效 × 计算成本"三维量化评估， 生成 Mermaid 过程追溯图，LLM-as-Judge 自动打分并输出靶向归因报告。自动触发：每次涉及 Skill 的任务完成后 自动采集执行数据并评分。触发信号：评测 skill、评估技能、skill evaluation、检查这个skill的质量、 这个skill怎么样、跑一下skill评测、skill质量报告。

jorinyang Bundle

双速技能进化引擎——吸收 MOMO CODE / Pioneer Agent 方法论，实现 Hermes 技能的自动观测→蒸馏→注入→固化闭环。 Fast Loop（/evolve 手动触发，秒-分级）：扫描信号→蒸馏候选 patch→Thompson 采样选择→注入预览→人工确认。 Slow Loop（Cron 30min 自动，时-日级）：聚合信号→skill-evaluator 对比→Ratchet Gate 只保留提升版本→飞书提案。 触发词：/evolve / 技能进化 / 进化技能 / 优化技能 / 看看哪些技能该优化了 / 技能自进化

jorinyang Skill

贵州之客智能行程规划 — 输入目的地/天数/偏好/预算/日期，调用 weather + amap + minimax 三重 MCP，产出结构化行程 + 地图标注 + 费用预估，双版输出（客户版 + 地接版）。

jorinyang Bundle

Agent 工具系统设计与生成工具——defineTool→registry→toolsToAI 三层架构方法论， 含脚手架生成器 + 合规章检查 + MCP Schema 导出。触发词：Agent 工具/工具系统/ 生成工具定义/defineTool/生成 defineTool/工具注册表/验证工具集/检查工具合规/ 导出 MCP Schema/toolsToAI/MCP tools 设计/Agent 工具箱/脚手架/工具骨架。

jorinyang Bundle

Darwin Skill Cron 自动化模式 — 技能每日自主优化：9维评分 + 棘轮机制 + 并行子 agent 执行。含自包含 prompt 模板和两日实战数据。触发词：darwin cron、技能自动巡检、skill nightly optimize。

orcaqubits Skill

Build an AP2 Merchant Payment Processor — the agent that constructs payment authorization messages, requests credentials from the Credentials Provider, processes payments, and returns receipts. Use when implementing the MPP role.

orcaqubits Skill

Implement the AP2 Agent Authorization Framework — open vs closed mandates, the two-phase delegation/authorization process, the cryptographic authorization chain, Key Binding JWTs, and human-present vs human-not-present signing. Use when designing how a user authorizes an agent to act, before touching any specific mandate type.

forexgod21 Bundle

Use for greenfield feature design — translating a product idea into an architecture, data model, API contract, and phased delivery plan. Use when starting from zero, designing new systems, evaluating build-vs-buy, or choosing between architectural alternatives. Triggers on "design a feature for", "how should I build", "architect a system that", "what's the right approach for", or any request that starts from a blank page with no existing implementation to extend.

forexgod21 Bundle

Use this skill whenever the user needs to track, evaluate, or act on the reliability history of an AI agent or human-AI team member over time. Triggers when the user asks about agent reliability, performance history, trust scoring, or says things like "how do I know if I can trust this agent", "track agent performance over time", "build a reputation system for my agents", "this agent keeps making mistakes — how do I handle that", or "how should past behavior affect how much I delegate to an agent." Always activate this skill when the user needs a structured method for evaluating agent trustworthiness based on demonstrated performance history rather than assumed capability.

forexgod21 Bundle

Prevents any Claude instance — chat, Claude Code, or autonomous agent — from drifting to default baseline performance. Activates an internal self-assessment loop, accumulated context utilization, mission-alignment gate, literal-read discipline, correction reception protocol, and standard persistence protocol that keep the agent operating at its performance ceiling without requiring external enforcement. Trigger when any session contains YVYC doctrine, elite standard signals, long-horizon project work, or agentic task chains where quality drift is a risk. Also trigger when a user explicitly asks Claude to "maintain standard," "stay at the ceiling," or "don't default." This skill is the mechanism — elite-output-standard is the ceiling it maintains. Both should run together when available.

forexgod21 Bundle

Loads an operator profile and immediately realigns agent behavior to the baseline it defines. Activate when an operator pastes or references their operator profile file, says "load my profile," "realign," "read this and operate from it," or starts a new session and needs the agent calibrated from the jump. Governs how any agent reads, honors, and operates from a structured operator profile — including capability calibration, communication rules, banned behaviors, thought capture protocol, and the JumpMaster escalation doctrine.

forexgod21 Bundle

Activate whenever an agent encounters a failed tool call, a failed action, an unexpected result, or an error of any kind during task execution. Governs how many retries are permitted, what must change between attempts, when to escalate to the operator, and how failures are recorded. Prevents infinite retry loops, silent failure swallowing, and unstructured error handling.

forexgod21 Bundle

Activate whenever an agent is about to report a task as done, finished, complete, ready, or shipped. Governs what evidence must exist before a completion claim is permitted, how partial completion is reported, and what an agent must never assert without verification. Prevents the single most damaging agent failure mode: confident completion claims for work that was never verified.

forexgod21 Bundle

Activate whenever the task involves designing, documenting, evaluating, or explaining how a program produces change — logic models, theories of change, program design, impact measurement, evaluation planning, outcome frameworks, or "how do we show this works." Trigger on any nonprofit or community program being built or assessed, any funder request for a logic model or theory of change, and any program that is running on faith instead of a stated causal chain. Fire at design time when possible — a logic model built for the funder after the program exists is a diagram; built before, it is a blueprint.

forexgod21 Bundle

Activate whenever the task involves designing how multiple AI agents, tools, and surfaces map to human roles and organizational functions — one person augmented by an agent stack, a team distributing work across AI systems, deciding which agent owns which lane, where authority lives and where it delegates, or "how should I structure my AI setup." Trigger on multi-agent architecture, agent role definition, human-AI org charts, and any operation consolidating multiple functions into fewer humans via AI augmentation. Fire at design time — agent stacks assembled by accretion develop overlapping lanes, orphaned functions, and authority ambiguity that surfaces as chaos precisely when the stakes are highest.

forexgod21 Bundle

Activate whenever a long-running AI agent compresses, summarizes, or compacts its context — long-horizon agent design, context compaction or summarization pipelines, multi-hour or multi-day agent sessions, memory consolidation systems, or any agent whose safety rules, boundaries, and standing instructions live inside a context window that gets rewritten over time. Trigger on designing compaction strategies, diagnosing agents that "forget their rules" mid-mission, or reviewing long-horizon deployments for constraint drift. Fire because compaction optimizes for task-relevant content — and safety constraints, which are not task content, are precisely what summarization silently drops.

forexgod21 Bundle

Activate whenever an AI agent's work is challenged, criticized, or found defective — pushback on an answer, a bug in delivered code, a flawed analysis, a wrong claim — and the response to that failure is what gets decided. Trigger on designing error-response behavior for agents, on any moment an agent is tempted to abandon its work under pressure, and on diagnosing agents that fold, over-apologize, or silently swap answers when questioned. Fire because the two default failure responses — capitulating to any pushback, and making the mistake disappear instead of fixing it — both destroy the thing the user actually needs: a reliable account of what is true and what broke.