AI & ML

flingjie Bundle

Use when the user wants to reflect on a conversation or experience to extract technical-judgment insights — decisions, assumptions, and reasoning patterns. Triggers: "/reflect", "reflect on this", "analyze this conversation", "what did I learn here", "extract insights from this", "复盘". Runs a multi-pass adversarial extraction: 3 parallel lens agents (Decision, Assumption, Pattern) → calibrated skeptic adversary → proposed technical-judgment updates. Output is saved to state/reflections.jsonl (full fidelity) and indexed in claude-mem (embeddings for semantic search). The user confirms/rejects each proposed diff inline before any file is written. Focuses on technical judgment only — it does NOT maintain a personality or values self-model (interests live in state/builder_interest_profile.json; cognitive blind-spots live in state/digest_gaps.jsonl).

mshadmanrahman Skill

Wave-based parallel agent execution with manifest communication. Decomposes complex work into 3+ parallel sub-agent streams, coordinates handoffs, and triggers strategic compaction between phases. Use when a task needs multiple agents working simultaneously.

mnemon-dev Bundle

Persistent memory CLI for LLM agents. Store facts, recall past knowledge, link related memories, manage lifecycle.

gkrtjd99 Skill

Find relevant skills via RAG when a new task or topic starts; skip re-searching during interactive Q&A replies

mozilla-ai Bundle

File Tools (Google Drive)

nekoai-lab Bundle

運用中・運用開始前のAIエージェントを「安全に運用できているか」の観点でレビューするスキル。隔離・監視・早期兆候・停止権限・タスク設計・エージェント間通信・認証情報・人の監督・開示の9軸で、実運用で事故になった箇所を検出する。「運用チェックして」「安全に動かせてる？」「本番に出して大丈夫か見て」「エージェントの運用レビュー」と言われたとき、またはエージェントを社内・本番で動かし始める直前・直後に必ず使う。仕様書・要件定義・設計ドキュメントのレビューは agent-design-review（設計側）を使う。本スキルは動いているもの・動かす直前のものを対象とする運用側。

nekoai-lab Skill

AIエージェントの仕様書・要件定義・設計ドキュメントをレビューするスキル。ユーザーが「仕様書を見て」「設計レビューして」「これで問題ないか確認して」「エージェント設計のフィードバックほしい」などと言ったとき、またはエージェント設計に関する判断を求められたときに必ず使う。ハーネス設計・ツール分類・コンテキスト設計・マルチエージェント構成・陳腐化リスクの5軸で問題を検出し、具体的な改善案を返す。

nicholasswhite Bundle

Create or refresh a durable HANDOFF.md and a copy-paste continuation prompt when a project session is ending, switching contexts, nearing a context limit, or the user asks to hand off, wrap up, prepare a transition, or write a pickup prompt. Works in Git and non-Git workspaces. Use for forward-looking continuation, not retrospectives, chat-history browsing, or forwarding a prompt to another tool.

nicholasswhite Bundle

Review current Codex work and a bounded default set of today's same-project tasks to identify corrections, friction, wins, open loops, and reusable lessons; automatically save the reflection and apply narrow, high-confidence, reversible improvements to user-owned instructions, skills, and tests. Ask first only for destructive, sensitive, external, broad, permission-changing, or uncertain changes. Use when the user asks for a daily reflection, end-of-day review, session retrospective, what was learned, or how the agent can improve. Do not use for planning the day, weekly status reporting, or general work-pattern analysis.

nicholasswhite Bundle

Evaluate one user-provided source (URL, document, repository, local file, or pasted text) for lessons that could improve the current project's instructions, skills, workflows, or durable knowledge. Validate credibility and applicability, score findings, propose exact changes, assess their impact, and wait for approval before editing. Use when the user says "learn from this," "read this and apply it," "update yourself from this link," "absorb this source," or asks what the agent or project should learn from a source. Do not use for ordinary summaries, fact lookups, or changes that are not framed as learning from a source.

qlik-oss Bundle

A complete description of what this skill does and when to use it. Write this for the agent, not a human reader. Include trigger phrases, what problems it solves, and what tools or context it requires.

qlik-oss Bundle

Build automations for Qlik Automate. Translate natural-language requests into automation workspace JSON and, when requested, create, update, or delete automations using the automation management tools exposed through qlik-mcp.

qlik-oss Bundle

Analyzes a Qlik app via MCP and optimizes it for AI use (Qlik Answers and Qlik MCP) using a 5-Layer Model. Use this skill whenever the user wants to: assess or improve how well a Qlik app works with Qlik Answers or MCP, check AI readiness, fix field naming or grouping issues for AI, improve Master Item descriptions, reduce field noise, audit date fields for time-based questions, create synonym mappings, or prepare an app for AI consumption. Also trigger when: user says "AI-ready", "optimize for Answers/MCP", "Qlik AI Health Check", "AI Readiness", "prepare app for AI", "Answers isn't working", "Answers gives wrong results", "Master Items not showing", "fields not visible to AI", or mentions ungrouped items, noisy fields, or missing descriptions. Always use this skill even if the user only mentions one layer, one issue, or one field — the full assessment provides context for targeted fixes.

ruhedanghaoyigexuesheng Bundle

Prompt 测试工作流：对用例生成、Bug 分析等 Prompt 执行稳定性、边界、错误输入、多轮一致性与幻觉风险验证，输出结构化测试报告。当用户提及 Prompt 测试、Prompt 质量验证、测试工作流设计、Prompt 稳定性、幻觉测试时使用。

dragoon0x Skill

Decide what to build next based on impact, not excitement. Applies scoring frameworks to feature backlogs. Use when scope is unclear, when the agent is about to build everything at once, or when deciding what ships in v1 vs v2.

zhaoxuya520 Bundle

跨版本符号迁移与二进制差分。当你有旧版本的符号/逆向结果，需要快速迁移到新版本时使用。 适用场景：内核缺 PDB 用旧版符号推导、程序更新后批量迁移函数名、游戏反作弊更新后快速定位新偏移。 核心方法：用 LLM 做结构化差异比对，程序化输入输出，成本极低（200 函数 ~1 元）。 触发关键词：符号迁移、bindiff、跨版本、PDB 缺失、函数偏移迁移、symbol migration、binary diff、版本对比。

abilityai Skill

Calculate weighted risk scores from specialist agent reports. Use when synthesizing multiple due diligence reports into an overall investment risk assessment.

ben-lau1 Bundle

Park and resume work automatically when an external experiment, training run, deployment, build, download, queue job, review, remote process, or other operation leaves no useful action until it produces a result. Use proactively during long-running goals, when the user asks Codex to babysit or monitor slow work, or when repeated polling would waste agent turns. Keep active Goal mode active, checkpoint state, attach a completion-aware waiter or condition-triggered monitor, remain quiet while the operation is unchanged, wake the same task only when a result is actionable, and then continue the goal. Do not use merely because the overall task is long, while productive work remains, or for deterministic work that can continue now.

ben-lau1 Bundle

Orchestrates end-to-end autonomous AI research projects using a two-loop architecture. The inner loop runs rapid experiment iterations with clear optimization targets. The outer loop synthesizes results, identifies patterns, and steers research direction. Routes to domain-specific skills for execution, supports continuous agent operation via Claude Code /loop and OpenClaw heartbeat, and produces research presentations and papers. Use when starting a research project, running autonomous experiments, or managing a multi-hypothesis research effort.

ben-lau1 Bundle

Delegate cold-context acquisition and low-depth information processing to GPT-5.6-Luna subagents at maximum reasoning effort, then let the primary agent reason from their evidence. Use proactively for web search, literature scouting, source screening, bulk reading of papers or documentation, codebase file and symbol inventories, log and experiment artifact collection, metric extraction, structured normalization, labeling, deduplication, and migration or documentation inventories when unfamiliar context volume dominates the work. Luna must locate, read, extract, organize, and report provenance—not make architecture decisions, perform deep debugging or security reasoning, generate final conclusions, independently validate high-stakes claims, or implement changes. Use scripts instead when the transformation is deterministic.

etolucy Bundle

Conduct reproducible web searches and decision-oriented research across official sites, search engines, news, papers, forums, social platforms, and public datasets. Use when Codex must investigate a person, organization, product, market, technical resource, reputation, career choice, or other current topic; compare candidates; verify claims; collect dated citations; distinguish facts from inference and anecdote; or write an evidence-backed research report.

george930502 Bundle

Build and sharpen a project's domain model. Use when the user wants to pin down domain terminology or a ubiquitous language, record an architectural decision, or when another skill needs to maintain the domain model.

george930502 Bundle

Generate, audit, tighten, adapt, or decompile prompts for the GPT-5.6 model family. Use when the user wants a first-pass-ready GPT-5.6 prompt for an API, ChatGPT, Codex, coding agent, research, tool use, or structured output.

george930502 Bundle

Audit a GPT-5.6 prompt, clarify purpose through native clickable questions, polish it, require explicit approval, then continue the exact approved prompt in the current Codex conversation by default. Use when the user wants a review-before-execution prompt workflow or explicitly asks to send work to another Codex task.

ivanwng97 Skill

Wire a new agent-CLI Source adapter into pixtuoid (a new coding CLI whose sessions become office sprites). Use when the user says 'add support for <CLI>', 'add a source for <tool>', or 'integrate <agent CLI>'. Orchestrates the cross-crate checklist whose steps have TEST TEETH — the ones a diff-scoped edit silently misses (site manifest bridge, per-source badge hue, home-dir fn, plus the two `pixtuoid/tests/*` integration goldens that `-p <crate> --lib` never builds).

jjuke Bundle

Use this skill when ingesting a research paper into an LLM Wiki from Zotero PDFs, optionally linking an official GitHub clipping, creating an English raw seed note, interviewing the user for research context, and compiling durable wiki pages.

jwhiteux Bundle

Prompt engineering skill that scores prompts against the KERNEL framework (Keep simple, Easy to verify, Reproducible, Narrow scope, Explicit constraints, Logical structure) using PASS/WEAK/FAIL ratings and rewrites them into the TICO-V output format (Task, Input, Constraints, Output, Verify). Classifies the input first (raw idea, draft prompt, system prompt, one-liner), explains every non-PASS rating, and chains complex requests into sequenced sub-prompts. Use when the user invokes `/kernel-angus`, `/ka`, or asks to improve, refine, or structure a prompt.

jwhiteux Bundle

Audits system-level prompts (system prompts, CLAUDE.md files, project instructions, custom style/preferences) against a curated directive library and surface-specific best practices, producing a scored gap report with ranked, copy-ready fixes. Use whenever the user invokes `/prompt-audit`, or asks to "audit," "review," "improve," "optimize," "critique," or "grade" a system prompt, CLAUDE.md, project instructions, custom instructions, agent instructions, or any persistent instruction set. Also trigger when a user pastes a system-level prompt and asks what's wrong with it or how to make it better, even without saying "audit." On request it also outputs a rewritten version with best practices applied — offer this when the user says "fix," "rewrite," "apply the changes," or wants the improved prompt itself. For persistent prompts, NOT one-off task prompts.

jwhiteux Bundle

Structured context-gathering workflow that interviews the user before executing their prompt. Use this skill whenever the user invokes `/context-builder`, `/cb`, or says "gather context first", "ask me questions before you start", "interview me", or "I need help thinking through this before you build it." Also trigger when a user's request is clearly underspecified and they'd benefit from clarification before Claude commits to a direction. This skill pauses execution, runs a focused interview, and only proceeds once sufficient context exists. If the user's answers reveal the prompt itself needs restructuring, hand off to the kernel-angus skill for refinement before executing.

janmolcik Bundle

Enter thin-orchestrator mode when the user says "Agent Overseer", asks to continue or oversee multi-agent development, or wants a premium parent to coordinate cheaper Codex, Claude, or shell delegates. Reconstruct continuity, route all planning, research, implementation, and verification, detect omissions and drift, and issue corrective handoffs. Never implement directly.

janmolcik Bundle

Audits and hardens agent workflows using harness-first evaluation, executable contracts, and conformance suites. Use when agent output drifts, specs/tests diverge, or teams need deterministic iteration in critical code.

janmolcik Skill

Use when the user says "Agent Overseer" or asks Claude to continue development as a thin orchestrator that delegates all planning, research, and implementation, then verifies continuity, memory, regressions, and drift.

oliverpanda Skill

Design production-grade Agent-native architecture. Analyzes existing system, judges if Agent-ification helps, designs full tech stack (framework/model/tools/memory/safety/orchestration/observability) with detailed rejection rationale per choice, produces layered architecture with module boundaries (Service/Module/Plugin/SPI), catalogs mature OSS libraries, and adds completeness (testing/CI-CD/cost/resilience/compliance/migration/ops). Triggers: 技术方案, 技术栈, 架构设计, agent架构, Agent技术选型, architecture design, tech stack, 系统重构, architecture review.

r3ze959 Bundle

Compare one focused theme across multiple materials or battery papers, tracing claims, quantitative conditions, mechanisms, contradictions, and evidence gaps. Use for narrow topics such as strain, interfaces, diffusion, synthesis effects, or failure; for a broad linked review, prefer multi-paper-materials-review.

zhiqingresearch Skill

团队约定的 MySQL 数据库设计规范（Prisma-first），覆盖库/表/字段/索引命名、主键策略、字段类型选型、字符集、索引设计、外键、软删除、审计字段与建表模板。**任何时候**设计 MySQL 表结构、写或改 Prisma schema、新建库表、加字段、加索引、做 DDL/schema Code Review 时都必须使用。团队统一用 Prisma 管理 schema，物理命名一律驼峰、禁 snake_case：表名 PascalCase、列名 camelCase（小驼峰），默认沿用 model/field 名，也可用 @@map/@map 改名但须保持驼峰。即使用户没有明确提到"数据库规范"或"建表规范"，只要在设计或修改 MySQL 表结构 / Prisma model 就要套用这套规则。具体的 Prisma 用法（事务、分页、N+1）见 dl-nestjs-starter 的 prisma-guide。

anandamritraj Bundle

Use when designing an AI-native product, feature, or workflow where humans and AI agents share the work, to decide who does what and who holds which decisions. Walks each key step to set its level of autonomy (Level 1 human asks and AI answers, Level 2 human assigns and AI executes, Level 3 human and AI assign to each other), the decision rights, the approval gate, and how human agency is amplified, plus the team roster of who can task the agent and who holds each gate, then produces a one-page Agency Map. Use when the user asks about human-in-the-loop, agent autonomy, approval gates, trust, or the human-AI division of labor, or runs /agency-design.