AI & ML

scouzi1966 Skill

Use when testing tool call reliability between OpenCode and afm — captures streaming XML tool call errors, classifies them as afm translation bugs vs model generation errors, and produces a diagnostic report without fixing anything

soujava Bundle

SDD4J architecture adapter for Java projects organized by technical layers such as controller, service, repository, model, domain, or dto. Use with SDD4J and a Java stack skill when specs need to map capabilities to classes spread across layer packages. Trigger on sdd4j-package-by-layer, package-by-layer, layered architecture, controller-service-repository layout, Spring-style layers, or SDD4J setup for existing layered Java projects.

soujava Bundle

SDD4J architecture adapter for Java projects organized by feature or capability package. Use with SDD4J and Java stack skills when each feature owns a package containing its spec, entrypoints, application code, domain/model code, and tests. Trigger on sdd4j-package-by-feature, package-by-feature, feature package, vertical slice package, capability package, or Java architecture mapping for co-located specs.

tommasobbianchi Skill

Build a map of an unfamiliar codebase BEFORE working inside it — porting, reimplementing, matching another implementation's behaviour, or any deep change. Use when about to port/translate/reimplement code, reproduce a reference implementation's output, or start substantial work in a repo you do not already model. Triggers on "port this", "reimplement", "match this implementation", "reproduce the reference", "traduci questo codice", "porta questo in", "riproduci il comportamento", "mappa il codice", "map this codebase", "understand this code first".

tommasobbianchi Skill

Metacognition — confidence calibration. Invocala prima di affermazioni, predizioni, riassunti, "ho fatto X funziona". Use to separate what you KNOW (observed/verified) from what you ASSUME (pattern matching, inference, "should work") — and to declare it explicitly to the user. Triggers anche quando stai per dire "credo", "dovrebbe", "probabilmente", "sicuramente", "verified".

azure Skill

Kars AGT E2E encryption skill — how the Signal Protocol inter-agent messaging works, how to debug it, and what was patched.

azure Skill

Spawn secure isolated sub-agent sandboxes, delegate tasks via AGT mesh, receive results, and destroy sub-agents. Uses the kars_spawn, kars_mesh_send, kars_mesh_inbox, and kars_spawn_destroy tools.

azure Skill

Behavioral governance for OpenClaw agents via AGT — tool-level policy, inter-agent trust, audit logging.

azure Skill

Query and inspect Foundry prompt agents and invoke Foundry tools via the Responses API. OpenClaw is the orchestrator — Foundry provides managed AI services.

azure Skill

Persistent long-term memory via Foundry Memory Store APIs. User preferences and chat summaries survive pod restarts — no Foundry hosted agent needed.

azure Skill

Knowledge retrieval (RAG) via Foundry file_search and azure_ai_search tools. Agentic retrieval with citations — uses Responses API.

azure Skill

Query model deployments, connections, and indexes in the Foundry project. Discover available models and infrastructure.

azure Skill

Evaluate agent quality using Foundry OpenAI Evals API. Create evaluations, run them against models, and analyze results.

mantasimb Bundle

Use only when the user explicitly invokes the BTW skill for a side-question branch. Reuse an existing read-only BTW sub-agent when available; otherwise create one with the current conversation and repository context, and return only a concise synthesized answer to the main thread.

mantasimb Bundle

Test-driven development with red-green-refactor and a mandatory CodeRabbit second-opinion review using `cr review --agent`. Use when the user wants to build features or fix bugs using TDD, mentions "red-green-refactor", wants integration tests, asks for test-first development, or wants CodeRabbit review integrated into the implementation workflow.

mantasimb Bundle

Interrogate product, architecture, implementation, and migration plans until the plan reaches a shared and operationally useful level of clarity. Use when a user wants Codex to stress-test a proposal, walk every branch of the design tree, resolve decision dependencies one-by-one, surface hidden assumptions, or turn a vague plan into a concrete execution model. Especially useful for RFCs, feature plans, refactors, system design notes, rollout plans, and codebase changes where some answers should be discovered by inspecting the repository instead of asking the user.

adelrioj Skill

Use BEFORE starting work on a task to surface the unknown-unknowns you're about to walk into but can't see. Dispatches a fresh-context Claude subagent that hunts what you don't know you don't know about the codebase, the domain, and the decisions ahead, calibrated to your stated experience, then reports ranked, evidence-backed findings and fixes nothing. Triggers on: blind spot pass, blindspot, unknown unknowns, what don't I know, what am I missing, help me prompt better, I know nothing about, teach me my unknowns, before I start, de-risk this task, surface my blind spots.

adelrioj Bundle

Use when a brainstorming design spec has been written and needs adversarial review before implementation planning, using a local model served by LMStudio (via the pi CLI) as the independent reviewer. Works offline — no hosted API needed. Triggers on: spec review local, local spec review, review spec locally, lmstudio review, review spec, spec review.

rengwu Bundle

Build a throwaway prototype to answer a design question. Use when the user wants to sanity-check whether a state model or logic feels right, or explore what a UI should look like.

rengwu Bundle

Plan a huge chunk of work — more than one agent session can hold — as a shared map of investigation tickets, and resolve them one at a time until the way to the destination is clear.

rengwu Skill

Compact the current conversation into a handoff document so a fresh agent can continue the work.

shaybush Bundle

Manage Cloudflare via the v4 API. Use whenever the user wants to work with their Cloudflare account — add a subdomain / DNS record, create or delete a domain (zone), create or list R2 buckets, add a secret to the Secrets Store, or verify their API token. Each operation is a script in scripts/ that loads the token from scripts/.env itself, so credentials are never read by the model.

relationalai Bundle

Build prediction task tables for predictive-modeling frameworks (e.g. GNN) that ingest Snowflake tables. Use this skill whenever the user wants to create, scaffold, or modify a task table for any of the six supported tasks — binary classification, multilabel classification, multiclass classification, regression, link prediction, or repeated link prediction. Trigger on phrases like "build a task table", "set up a prediction task", "label table", "training table for the GNN", "churn prediction task", "link prediction labels", and on any mention of task tables alongside Snowflake. Also trigger when the user has source tables and wants help defining a prediction problem against them, even if they don't yet know which of the six task types fits. Supports both a guided, step-by-step interview and a low-question one-shot mode — this skill asks which one up front. Not for building the graph/model from an existing task table — see `rai-predictive-modeling` — or for training/predictions/evaluation — see `rai-predictive

stackloklabs Skill

Guide for developing MCP servers with Python SDK. Use when implementing MCP tools, resources, or prompts.

alnitak Skill

Core mental model for the flutter_soloud audio plugin — the SoLoud.instance singleton lifecycle (init/deinit), the AudioSource vs SoundHandle distinction, synchronous play semantics, and cross-platform constraints. Use when the user asks to play a sound, set up audio in a Flutter app, debug SoLoudNotInitializedException or silent playback, or migrate from audioplayers/just_audio.

alnitak Bundle

Teaches the push buffer-stream API of the flutter_soloud package (setBufferStream, addAudioDataStream, setDataIsEnded, BufferingType, BufferType, icy metadata). Use when the user wants to play audio that arrives in chunks — internet/icecast radio, WebSocket PCM feeds, TTS/LLM streaming APIs, or procedurally generated PCM — instead of loading a complete file or asset.

alnitak Skill

Teaches volume, pan, fade, oscillator, and scheduled-stop control in the flutter_soloud package, including the handle-based model, setPanAbsolute, getApproximateVolume for VU meters, and the limiter filter as the anti-clipping tool. Use when the user asks to change a sound's volume, pan a sound left/right, fade or oscillate audio parameters, build a VU meter, or stop clipping when many sounds play at once.

avisheku Bundle

Clarify, research, confirm ShortPlan, model-aware implement, 2080, handoff.

avisheku Bundle

Switch MCP profile explicitly; restore minimal after work.

danielteles Bundle

Modern Angular conventions for writing, reviewing, refactoring, or migrating Angular: standalone components, `OnPush` and signals, `input()` / `output()` / `model()`, `@if` / `@for` / `@switch` control flow, `inject()` and functional providers, functional guards and interceptors, Signal Forms and typed reactive forms, lazy routing, zoneless-ready change detection and SSR, testing by role. Builds on `core-typescript` and `architecture-and-design`. Use it when the user mentions Angular, signals, `computed`, `effect`, standalone, `OnPush`, control flow, `inject`, RxJS, reactive forms, change detection, zoneless, `NgModule`, or `ng generate`.

danmaps Bundle

Generate ArcPy scripts for ArcGIS Pro using live map context, layer sampling, and ArcGIS‑aware prompt rules

danmaps Skill

Decide if a dataset/project is ready for a specific spatial analysis (buffer, overlay, join, routing) and list blockers + preflight steps.

fedoroff-vlad Skill

Use when creating a new module from the project's canonical layout — a new service, MCP module, agent, or migration. Fires on: "add a new module", "scaffold a service", "new MCP module", "new agent", "new migration". Copies the canonical layout, wires it in, and runs check-drift so no coupled artifact is forgotten.

fedoroff-vlad Skill

Use after ANY change and before opening a PR to verify every coupled artifact moved together. Fires on: changed an env var / port / endpoint / MCP tool / dependency / LLM model / DB schema; "did I update everything"; "check drift"; "reconcile the docs". Reads the repo's .skills/change-map.yaml and reports which coupled files did NOT move.

fedoroff-vlad Skill

Use when you want to run the golden LLM tests against a real model — one, a few, or all. Fires on: "прогони голдены" / "run the goldens", "прогони golden тесты", "проверь роутинг/синтез на реальной модели", "did my prompt/skill/router change break the goldens", after editing a prompt, SKILL.md, classifier, or any LLM surface. Discovers the repo's golden tests, lets you pick a scope, runs them through the repo's golden runner, and reads the result (real regression vs flaky borderline case).

fedoroff-vlad Bundle

Use after the project is mapped, to write the human- and agent-facing docs: a per-module README, a root README with the module map, and an AGENTS.md spec skeleton so an agent can develop against it. Fires on: "сделай ридми по проекту", "напиши README по каждому модулю", "задокументируй проект", "нужна документация для онбординга", "write the READMEs", "document each module". Reads docs/onboarding/project-map.md; writes README.md per module + at the root, and a spec skeleton.