AI & ML

mayank9mare Bundle

Use when building, reviewing or debugging an LLM agent, tool-calling loop, or AI workflow — choosing between a chain, router, evaluator-optimiser or agent loop; designing tools and their schemas; multi-agent orchestration; RAG retrieval quality; prompt caching and token cost; adding tracing with OpenTelemetry GenAI conventions; evaluation and LLM-as-judge; prompt injection and tool permissions; human-in-the-loop approvals; durable/resumable runs. Also for "my agent loops forever", "the agent picks the wrong tool", "LLM costs too much", "how do I trace an agent", "agent gives inconsistent answers".

mayank9mare Bundle

Use when building, scaffolding, or reviewing a Python backend service or API — choosing between FastAPI, Litestar, Django/DRF, django-ninja, Flask, Starlette or FastMCP for MCP servers; SQLAlchemy 2.0 async sessions and Alembic; Pydantic v2 and pydantic-settings; Redis caching; Celery/arq/Dramatiq workers; httpx timeouts; uvicorn/gunicorn/granian worker config; uv packaging; structlog and OpenTelemetry; pytest-asyncio and testcontainers; Dockerfiles for Python. Also for "start a Python service", "build an MCP server", "why is my FastAPI slow", "blocking the event loop", "async SQLAlchemy setup".

maggielerman Bundle

Custom skill created by Maggie Lerman. Generate, analyze, validate, preview, and safely promote a Hyphenomenon Codex usage report. Use for daily or weekly Codex activity reports that require deterministic telemetry, GPT-5.6 Sol/high interpretation, Agent Stack evidence, Chronicle context, and fail-closed publication gates.

jajabong Skill

Author in-repo SKILL.md files: frontmatter and structure.

sidsivakumar67 Bundle

Structured adversarial review of a proposed plan or solution using two blind Sonnet-tier subagents (Angel and Demon) with opposing charters, weighed by the main model. Manual-only — invoke explicitly as /angeldemonprotocol or /angeldemonprotocol(n).

maggielerman Bundle

Use when asked to set up, use, triage, or maintain Context Layer UX/UI bug intake for user, agent, or subagent-detected UI defects.

igoakulov Bundle

Manage a local knowledge vault as AI agent. Ingest sources (ChatGPT, X, other), create/update topics from synthesized sources, organize coherent graph, find knowledge on demand, run health checks. CLI provides deterministic vault operations and token-efficient graph traversal. Use for backing up knowledge, building persistent context between owner/user/agent, or tracking learning progress.

ardha-eco-system Skill

Adversarial verification of finished work. Treats any 'done' claim as a set of claims to verify by re-running everything yourself. Use when someone reports work complete and you must gate it: load this before trusting a completion report. Subcommands: suite <target> runs a trap suite against any skill/model.

ardha-eco-system Bundle

A 7-step problem-solving discipline loop that gives any model structured thinking: classify the ask, define done, gather evidence, decide, act surgically, verify by observation, report outcome-first. Use when the user says 'fable-method', or proactively when starting any multi-step task that no task-specific skill covers. Subcommands: plan, audit, report.

maggielerman Bundle

Custom skill created by Maggie Lerman. Split repository work into dependency-safe parallel clusters, then bootstrap branches/worktrees, GitHub tracking issues, and ready-to-run agent prompts. Use when a user asks to run multiple agents in parallel or wants branch/worktree cluster setup.

maggielerman Skill

MUST be used first when prompts mention Metafields or Metaobjects. Use Metafields and Metaobjects to model and store custom data for your app. Metafields extend built-in Shopify data types like products or customers, Metaobjects are custom data types that can be used to store bespoke data structures. Metafield and Metaobject definitions provide a schema and configuration for values to follow.

brashler Skill

Add simple 3D sphere physics (gravity, bouncing, collisions) to three.js / 3js scenes, one-shot browser games, and demos using the Flit engine. Use when a task involves lightweight game physics, bouncing balls, particle simulations, collision detection, or a three.js scene that needs physics without a heavy engine. Includes a no-code MCP server option.

docxology Bundle

Compile selected public website pages into portable agent skills with Skillarum; inspect provenance, refresh source inventories, validate packages, and rebuild the evidence-bound manuscript. Use for Skillarum workflows and generated-skill maintenance.

isabellayoungbrown52-hash Bundle

Operate COMSOL from natural language through Java/Method batch or an MCP session: build, solve, debug, postprocess, interpret, and deliver a complete directly viewable MPH with model and result trees. Use for practical COMSOL modeling, sweeps, result extraction, visualization, and physics-aware interpretation. Choose the lightest workflow that advances the task; reserve full evidence gates for formal reproduction, promotion, or disputed results.

openmodelingfoundation Bundle

Use this skill when planning or reviewing FAIR stewardship of digital research objects, including research software, datasets, computational models, and workflows. Use it for FAIR metadata, reproducibility assessment, object/workflow provenance, persistent identifiers, citation, repository organization, dependency and environment management, packaging, portability, archival preparation, preservation, stewardship artifacts, and management planning. FAIR owns stewardship metadata, reproducibility assessment, object/workflow provenance, packaging, and pointers to scientific artifacts; it does not author or canonicalize model-card content. Triggers include making research objects FAIR, reproducible, citable, reusable, publication- or archive-ready; creating FAIR metadata (e.g. codemeta.json, CITATION.cff, RO-Crate); packaging or archiving a repository, dataset, or model; and developing FAIR, data, software, or maintenance plans. Expected output: a FAIR Management Plan as the sole canonical stewardship document, w

maggielerman Bundle

Act as the primary project manager and agent orchestrator for repos using the standard docs scaffold. Use when the user wants one primary agent to read project docs, derive the next tranche, delegate bounded work to subagents, manage handoffs, review completion, and write status updates back into the existing docs without creating a second source of truth.

openmodelingfoundation Bundle

Use this skill when planning or reviewing the implementation of a computational model as research software. Helps translate scientific models into maintainable implementations, preserve traceability to the conceptual model, identify implementation risks, and delegate platform-specific practices to specialized guidance. Expected output: implementation guidance, implementation review, implementation planning, and routing to platform-specific implementation guidance.

openmodelingfoundation Bundle

Generate narrative documentation for computational models and research software, and select the appropriate documentation framework for a given model type. Use this skill when a user wants to: - document a computational model - generate documentation from source code - write an ODD or ODD+2 narrative for an agent-based model - create model narratives for publication or reuse - draft narrative workflow documentation The skill classifies the model type, selects a framework, extracts model structure from supplied materials, and drafts documentation. It does not assess or score existing documentation — use the document-review skill for completeness assessment, gap analysis, or structured review output. For model-card requests or scientific model specifications, route to omfa. Inputs may include source code, pseudocode, READMEs, publications, architecture descriptions, or model metadata.

openmodelingfoundation Bundle

Evaluate computational model submissions for OMF peer review readiness using reproducibility, documentation, code quality, and research software engineering criteria. Use this skill whenever a user asks to review a computational model, codebase, model release, or submission package for publication, reuse, or peer review. Trigger on phrases like: "peer review my model", "is this model submission ready", "review codebase quality", "check reproducibility", "review ODD documentation", "assess FAIR/research software quality". Expected output: structured peer review report with a binary recommendation limited to named baseline submission criteria, criterion-by-criterion findings, prioritized fixes, and an evidence-based checklist mapped to those criteria and key EVERSE research software quality indicators.

maggielerman Bundle

Build your app's primary user interface embedded in the Shopify admin. If the prompt just mentions `Polaris` and you can't tell based off of the context what API they meant, assume they meant this API.

lzh20060217 Bundle

Run a gated mathematical-modeling competition workflow covering contest rules, problem and data decomposition, literature evidence, preprocessing and model approval, reproducible implementation, independent sensitivity and robustness validation, paper-outline handoff, and user-commanded packaging. Do not silently change a frozen plan or package without an explicit command.

7df-lab Bundle

Install the agent skills into $DEVO_HOME/skills from a curated list or a GitHub repo path. Use when a user asks to list installable skills, install a curated skill, or install a skill from another repo (including private repos).

masonweb3 Bundle

出片魔改层——在开源镜头卡库 video-shotcraft 之上加你自己那一层（红线映射、四阶段两审核点、派工单、收货 QC）。当用户要把一次实测做成成片、要派 Agent 渲片、要收货验片时使用。框架＝Remotion。

maggielerman Bundle

Custom skill created by Maggie Lerman. Act as the primary project manager and agent orchestrator for repos using the standard docs scaffold. Use when the user wants one primary agent to read project docs, derive the next tranche, delegate bounded work to subagents, manage handoffs, review completion, and write status updates back into the existing docs without creating a second source of truth.

jajabong Skill

Update Hermes Agent and repair Node/npm version drift.

jajabong Bundle

Consume DAG/kanban/agent task events; dedupe and aggregate for Queen.

jajabong Bundle

Use, configure, theme, extend, and orchestrate Hermes Agent.

taotaoboom Bundle

访谈式方案制定——执行任务前与用户逐问逐答共同理清方案细节，避免误解与返工。This skill should be used when the user expresses co-planning intent (e.g. "访谈式", "每次问一个", "我们讨论一下", "先定方案", "一起搞") or when a task involves multi-step plan design, new features, architecture decisions, or ambiguous requirements and the agent proposes an interview-style walkthrough. Guides upfront recon with tools, strictly one-question-per-turn interviewing with option-based questions and trade-offs, dynamic re-planning on new facts, and a decision-summary confirmation gate before execution.

phattbeats Bundle

Interact with the Paperclip control plane API for task coordination and governance. Use when checking assignments, updating issue status, posting comments, delegating work, managing routines, or calling Paperclip API endpoints. Also use when triaging an OCPlatform gateway adapter recovery wake (agent status=running with errorReason="OCPlatform gateway run failed") on a PHATT TECH agent — see OCPlatform Gateway Adapter section below.

phattbeats Bundle

Portable Paperclip board client — connect to any Paperclip (paperclipai) control plane from any agent harness with nothing but curl. Create tasks/issues, nest and block them, label them, attach files and documents, comment, assign and wake agents, schedule routines, handle approvals, and monitor the org — including agent cost and subscription quota usage — over REST. Use when the user wants to create a task on their Paperclip board, plan work as an epic with subtasks, hand deliverables to their agent org, find what the org has produced, check task or org status, see how much of a provider plan the agents have burned, or wire an external session (Cowork, Claude Code, cron, CI) into Paperclip. Trigger on: "paperclip", "portaclip", "put this on the board", "hand this to my agents", "create a task for the org", "wake the agent", "what are my agents working on", "what did the org produce".

phattbeats Bundle

Generate a working CLI from any API, then wrap it in an OpenClaw skill. Point it at API docs, a live URL, or manual endpoint list and get a dual-mode Commander.js CLI (human + agent JSON output) plus a ready-to-use skill folder. Use when user wants to wrap an API in a CLI, generate a CLI from API docs, turn an API into a command-line tool, scaffold a CLI from discovered endpoints, or create a skill for an API.

phattbeats Bundle

Crawl and extract content from any website using Crawl4AI. Returns clean markdown, metadata, links, and structured data from web pages. Use when an agent needs to research a website, extract page content, scrape business listings, analyze competitor sites, get clean readable text from a URL, take screenshots of web pages, or generate PDFs. Handles JavaScript-rendered pages. Preferred over web_fetch for rich extraction and over browser automation for simple content retrieval.

phattbeats Bundle

Use PHATT Blender confidently: free commercial assets, safe project loading, verified previews and GPU renders.

alphaparkinc Bundle

Agent runtime channel router and teammate dispatcher connecting self-hosted OpenClaw gateways to GenTeam multi-agent swarms.

rogerchappel Bundle

Prompt Regression Skill

noncelogic Bundle

Run coding tasks as persistent, detached jobs using the Claude Agent SDK. Jobs survive agent timeouts, gateway restarts, and API hangs.