AI & ML

bk202503 Skill

When the user has N independent improvements against a single upstream repo and each unit must land as its own upstream PR ("각각 PR로 올려", "open a PR for each", "fan out and send N PRs", "split these findings into separate PRs", "one PR per unit against upstream"), fan out N general-purpose subagents — each in its own git worktree with non-overlapping file scope — where every agent commits, pushes, and runs `gh pr create --base <base>` against the upstream repo, then the parent reports the N PR URLs. Distinct from parallel-dev (which merges locally into the base branch and is right for solo/personal repos) and parallel-dispatch (which is read-only status/lookup fan-out with no writes and no worktrees).

cor-incorporated Bundle

Generate an agent-to-agent handover document (経路B) or next-session handoff from the current session state: repo, worktree, branch, open PRs/issues, scope, constraints, and the mandatory rule block. Invoke with: /handover [宛先: codex|cursor|opencode|次セッション]. Use when the user says 'ハンドオーバー作って', '引き継ぎ文書', '次のセッションへの引き継ぎ', 'Codexに渡す資料', 'handoff', or when context is running low and work must continue elsewhere. Do NOT use for: progress reports (use /wave-report), PR descriptions, or delegating a small single task (write the task prompt directly).

j5ik2o Bundle

Use only when the user explicitly asks for Clean Architecture design, implementation, or review. Applies a domain, use case, interface adapter, and infrastructure layer model, with persistence and RPC adapters outside the domain and use-case rules. Do not trigger for generic architecture advice, hexagonal architecture, onion architecture, or ordinary design review unless Clean Architecture is named.

j5ik2o Bundle

Use when designing or reviewing repositories in DDD. Treats repositories as collection-like access to aggregate roots, not generic DAO layers, and separates query/read-model concerns from command persistence. Trigger for repository interfaces, aggregate persistence, save/load semantics, query placement, ORM leakage, or repository anti-patterns.

j5ik2o Bundle

Use when explaining why CQRS often needs event sourcing or durable events for reliable command-to-query synchronization. Covers calculated value sync, database trigger limits, polling scalability, double commits, and making events the source of truth. Trigger for CQRS synchronization, read-model update strategy, CQRS without ES, or double-commit concerns.

j5ik2o Bundle

Use when planning a DDD implementation flow that starts from the domain model and tests before infrastructure. Covers test-first domain modeling, in-memory repositories, use cases, and delaying persistence/framework choices. Trigger for starting from the domain model, TDD with DDD, in-memory repository design, use-case tests, or DDD development process.

j5ik2o Bundle

Use for choosing and applying DDD building blocks: value objects, entities, aggregates, domain services, repositories, factories, and domain events. Trigger for domain modeling, entity versus value object decisions, aggregate boundaries, domain service placement, or implementing a domain model from requirements.

j5ik2o Bundle

Use when analyzing existing non-DDD or legacy code to propose DDD domain models with aggregates, local entities, value objects, domain services, invariants, and pseudocode. Trigger for extracting a domain model from code, redesigning legacy code with DDD, finding aggregates, or proposing domain boundaries from an existing implementation.

j5ik2o Bundle

Use when designing or reviewing domain-model accessors that expose internal state for persistence, serialization, tests, or framework integration. Applies the breachEncapsulationOf naming pattern so unavoidable getters are visibly treated as encapsulation breaches, not casual domain APIs. Trigger for getter naming conventions, persistence-only getters, serialization accessors, Tell Don't Ask pressure, or requests to prevent getter misuse.

j5ik2o Bundle

Use only for Scala 3 implementations of CQRS/Event Sourcing with Apache Pekko. Covers aggregate actors, PersistenceEffector-style command handling, domain-model separation, typed state transitions, event design, Protocol Buffers serialization, ZIO use cases, and read-model updaters. Do not trigger for non-Scala stacks or conceptual CQRS questions without Pekko/Scala implementation work.

j5ik2o Bundle

Use when a use case wants to check another aggregate's state before changing the current aggregate. Helps separate true invariants from policies, detect Saga misuse, reason about CQRS/Event Sourcing lookup limits, and decide whether inconsistent data is acceptable. Trigger for cross-aggregate validation, checking another aggregate, command-side read-model access, reverse lookup constraints, or Saga-based validation.

moltis-org Bundle

Guarantee valid JSON/XML/code structure during generation, use Pydantic models for type-safe outputs, support local models (Transformers, vLLM), and maximize inference speed with Outlines - dottxt.ai's structured generation library

moltis-org Bundle

Remove refusal behaviors from open-weight LLMs using OBLITERATUS — mechanistic interpretability techniques (diff-in-means, SVD, whitened SVD, LEACE, SAE decomposition, etc.) to excise guardrails while preserving reasoning. 9 CLI methods, 28 analysis modules, 116 model presets across 5 compute tiers, tournament evaluation, and telemetry-driven recommendations. Use when a user wants to uncensor, abliterate, or remove refusal from an LLM.

moltis-org Bundle

Foundation model for image segmentation with zero-shot transfer. Use when you need to segment any object in images using points, boxes, or masks as prompts, or automatically generate all object masks in an image.

xu-xiang Skill

选择在解析结构化文本时使用正则表达式还是大型语言模型的决策框架——从正则表达式开始，仅在低置信度的边缘情况下添加大型语言模型。

first-fluke Bundle

Multi-vendor AI image generation with authentication-aware parallel dispatch. Routes to Codex (gpt-image-2 via ChatGPT OAuth), Antigravity (Gemini-family "nano-banana" image models via `agy` CLI + Gemini Code Assist; exact model chosen internally by agy), and Pollinations (flux/zimage, free with signup). Use for image generation, image creation, visual asset generation, and AI art.

first-fluke Bundle

Short-form, explainer, and demo video generation via a key-optional 3-tier router. Composes scripts, oma-voice narration, oma-image/oma-slide/stock visuals, key-free captions, and a per-run agent-authored Remotion composition (always-latest Remotion + remotion-dev/skills) into reproducible run directories. Routes three modes — shorts/reels (9:16), explainer (16:9 README/code/data), and demo/walkthrough (screen capture, incl. supervised headed web-app capture of any URL). Use for video, shorts, reels, short-form, demo, explainer, walkthrough, screencast, web capture, video generation, 영상, 숏폼, 쇼츠, 릴스, 데모, 설명 영상.

first-fluke Bundle

Automated multi-agent orchestration that spawns CLI subagents in parallel, coordinates via MCP Memory, and monitors progress. Use for orchestration, parallel execution, and automated multi-agent workflows.

galihadiprayoga Bundle

Use when setting up a new project (or onboarding an existing repo) for safe AI-agent development — scaffolding the CLAUDE.md/AGENTS.md protocol plus DevSecOps guardrails. Triggers include "initial project setup", "scaffold our protocol", "set up the agent rules", or the command /init-project.

galihadiprayoga Skill

Laravel testing strategies with PHPUnit, Pest, model factories, HTTP tests, Sanctum authentication testing, mocking, and coverage.

galihadiprayoga Skill

Orchestrate multi-agent coding tasks via Claude DevFleet — plan projects, dispatch parallel agents in isolated worktrees, monitor progress, and read structured reports.

galihadiprayoga Skill

Use this skill when retrieving Jira tickets, analyzing requirements, updating ticket status, adding comments, or transitioning issues. Provides Jira API patterns via MCP or direct REST calls.

cor-incorporated Bundle

Plan and execute multi-agent team compositions for parallel implementation tasks. Provides team patterns, wave-based execution, cross-review protocols, and TeamCreate vs parallel Agent decision matrix. Use when: planning team composition for complex multi-file features, choosing orchestration strategy for 3+ parallel tasks, setting up wave-based execution with dependencies, coordinating cross-review between agents. Do NOT use for: single-agent tasks, basic parallel execution of 2 independent tasks (rules handle that), Codex CLI delegation, or sequential single-file changes.

cor-incorporated Bundle

Design features using DDD + TDD + Clean Architecture + Event-Driven + API-First patterns for TypeScript, Python, and Go. Use when: designing features with domain complexity (Entities, Value Objects, Aggregates), structuring projects with layer separation, defining Bounded Contexts, applying Outside-In TDD, designing CQRS or Event Sourcing, creating API contracts before implementation, starting a new project architecture. Trigger phrases: DDD, TDD, clean architecture, hexagonal, ports and adapters, CQRS, event sourcing, API-first, domain model, bounded context. Do NOT use for: simple scripts, one-off utilities, UI-only components without domain logic, or CRUD-only endpoints with no business rules.

natan-mohart Bundle

Builds a full amortization schedule for one or more debt tranches via a bundled calculator, computes debt service coverage ratio against a minimum threshold every year, and flags exactly which year DSCR would breach, before it happens rather than when the covenant is actually tested. Use whenever the user needs a debt or amortization schedule, wants to check debt service coverage against a covenant, or has debt modeled as a single interest expense line with no actual amortization mechanics or DSCR check behind it.

natan-mohart Bundle

Reconciles leadership's top-down target against the driver-based bottom-up build line by line via a bundled calculator, flagging every gap over 10% instead of quietly averaging the two into a number nobody actually believes. Use whenever the user is building the annual operating plan, needs to reconcile a leadership target with a bottoms-up model, or has a plan where the top-down ask and the bottom-up build were never actually compared line by line.

natan-mohart Bundle

Builds an integrated P&L, balance sheet, and cash flow statement from driver assumptions via a bundled calculator that runs real integrity checks every year, does the balance sheet balance and does cash on the balance sheet tie to the cash flow statement's closing balance, instead of trusting a model that was never actually checked. Use whenever the user needs a financial model that links all three statements, is building a multi-year projection, or has a model where the balance sheet was never actually verified to balance.

natan-mohart Skill

Runs the full day-1-to-diagnosis arc of a real strategy engagement — a MECE hypothesis tree built before any data is pulled, then three-lens fact triangulation that proves or kills each hypothesis, with every finding tagged fact/inference/unknown and a workplan naming who closes the open questions. Use this at the literal start of any strategy engagement, whenever leadership is debating solutions before agreeing on the problem, whenever a broad fuzzy question needs to become a structured testable plan, or whenever the user asks "what's actually going on" with a business, team, or metric.

natan-mohart Bundle

Builds a driver-based economic model — NPV, IRR, payback, and a real sensitivity grid — using a bundled calculator, then states the specific conditions required for the case to hold. Use whenever the user wants to build, check, or challenge a business case, ROI, or investment case, is deciding whether to fund an initiative, or has a business case that's just a single-point revenue projection with no sensitivity analysis.

natan-mohart Skill

Designs how an organization actually gets work done — structure, decision rights (RAPID/DACI), processes, and capabilities — aligned to the strategy it's supposed to execute, instead of redrawing an org chart in isolation. Use whenever the user wants to redesign org structure, is dealing with slow or unclear decision-making, is restructuring after a strategy change, or asks for a new org chart without having named who actually decides what.

abchoudh-amd Bundle

Report-only multi-agent review of the current local rocprofiler-compute branch: diffs committed, uncommitted, and untracked work against an explicit base, fans out seven Cursor Task reviewers, runs gates, verifies findings with three skeptics, and publishes one local Markdown report inline with its absolute path. Use for deep self-review, local branch audits, and pre-PR review.

abchoudh-amd Bundle

Profiles the rocprof-compute source CLI in the current worktree using Scalene's CPU-only mode, interprets CPU hotspots, and supports agent-led before/after comparison. Use when a user asks to profile rocprof-compute, optimize a subcommand (profile/analyze/etc.), find CPU bottlenecks, or compare optimization results.

abchoudh-amd Bundle

Creates an immutable continuation snapshot of active agent work, saved relative to the invocation directory. Use when a user asks to create an agent handoff, transfer work between agents or sessions, preserve context before stopping, or prepare resumption notes.

onfire-ai Skill

Run AI prospecting against a target company to rank the likely buyers/champions to approach at that ACCOUNT, using the Onfire `ai_prospecting` tool. Use when the user wants prospects for an account, a ranked list of who to talk to at a company, or a batch run across multiple companies, and wants the engine to decide who is best. NOT for deterministic title/persona/seniority/location/technology lookups (route those to `ask_onfire` over the contact entity) or single-person "is X worth a meeting" checks. Covers `action='run'`, the dataset-backed response and how to query it, the polling pattern for long runs, and the shadow-schema test mode.

zad111ak-ai Skill

Reflection + Self-Correction + Quality Assurance протокол. Stop-Reflect-Correct-Verify-Deliver перед ответом на сложные задачи. Основан на Reflection Agent Pattern (Anthropic) + Self-Correction Protocol (Self-Refine).

zad111ak-ai Skill

Структура и стиль ответов пользователю в Telegram. Минимизация итераций, максимум полезной информации с первого раза. Без AI-слога.