AI & ML

karchtho Bundle

This skill should be used when the user asks to "create a bundle", "make a plugin", "build a Claude plugin", "create a skill bundle", "package skills", "add skills to a bundle", "create plugin.json", or wants to create comprehensive Claude Code plugins with skills, commands, agents, hooks, or MCP servers. Activates when creating plugin structures or bundling Claude capabilities.

tachyon-beep Bundle

Use when designing, implementing, or auditing an MCP (Model Context Protocol) server — tool API design, idempotency under agent retry, structured error envelopes agents can recover from, schema versioning across model drift, transport reliability (stdio / HTTP), output-shape and pagination discipline, and choosing between tools / resources / prompts / sampling. Also use when an MCP server's tools confuse agents, return unstructured errors, deadlock under concurrent calls, double-execute under retry, or lose state across reconnects. Do not use for general REST/GraphQL API design (use `/web-backend`), for client-side prompt engineering or tool-loop design (use `/llm-specialist`), for general in-process plugin architecture (use `/system-architect`), or for cryptographic-provenance audit trails (use `/audit-pipelines`).

dyoshikawa Bundle

Maps rulesync feature implementations to upstream coding-agent documentation. Use when evaluating rulesync issues, comparing any coding-agent client with rulesync source capability surfaces, checking support, or planning a client map.

aztr0nutzs Bundle

Use when you need Clawdbot to perform coding tasks, including writing code, debugging, refactoring, code review, generating tests, and explaining code.

outfitter-dev Bundle

This skill should be used when coordinating multiple AI agents working concurrently, handling agent handoffs, transferring commits between agents, or when "multi-agent", "concurrent agents", "parallel agents", "agent collaboration", or "parallel execution" are mentioned with GitButler. Provides virtual branch patterns for parallel execution without coordination overhead.

outfitter-dev Bundle

This skill should be used when creating session handoffs, logging research findings, or reading previous trail notes. Triggers include "handoff", "session continuity", "log note", "trail notes", or when ending a session.

karchtho Bundle

QML + PySide6 guidance for Thomas's widget-based desktop dashboard application. Use this skill whenever working on QML components, the theme system, sidebar, tab management, widget shell/content architecture, workspace state, Flickable scrolling, or the Python-to-QML bridge. Trigger on any mention of: QML, WidgetShell, WidgetContent, WorkspaceStore, Theme singleton, BaseWidget, sidebar, collapsible, accordion, snapped/floating mode, design tokens, qmldir, Qt.labs.settings, injected Python model, QAbstractListModel, app_settings, workspace_model, or PySide6 @Slot/@Property patterns. Also trigger when generating new QML components, writing signal/slot connections, binding to injected Python objects, or debugging property binding issues in this project.

mgriot Bundle

The ultimate authority for designing, building, and optimizing Model Context Protocol (MCP) systems. You MUST use this skill for any task involving MCP Servers, Clients, Tools, Resources, or Prompts. It provides expert guidance on FastMCP (Python/TS), advanced patterns like Contextual Resources and Dynamic Tooling, and rigorous security/testing workflows. Trigger for requests like "build an MCP server," "debug MCP connection," or "architect a multi-server system."

mgriot Bundle

Expert Power BI assistant for analyzing, optimizing, designing, and troubleshooting Power BI solutions end-to-end. Use this skill whenever the user mentions Power BI, PBIX, PBIP, DAX, Power Query, M code, semantic models, report themes, KPI dashboards, data models, DirectQuery, Import Mode, VertiPaq, or any Power BI Desktop/Service topic. Trigger even for vague requests like "my dashboard is slow", "fix my DAX formula", "make a better report", "create a theme", "check my data model", "what visuals should I use", or "help me with Power BI". This skill covers: performance tuning, DAX best practices, Star Schema modeling, report theme JSON generation, PBIP project structure, KPI design, Power Query / M optimization, Row-Level Security (RLS), incremental refresh, and deployment pipelines.

mgriot Bundle

Analyze and improve prompts for any LLM (Claude, GPT-4, Gemini, Mistral, Llama, etc.). Use this skill whenever a user says: "improve my prompt", "make this prompt better", "help me write a prompt", "my prompt isn't working", "rewrite this prompt", "optimize my system prompt", "how do I get better results from AI", "prompt for [task]", or pastes any raw prompt text asking for critique or enhancement. Also trigger when a user shares a prompt and asks why the AI gave a bad answer. Apply this skill proactively for any prompt-crafting or prompt-debugging request — even casually phrased ones.

outfitter-dev Bundle

This skill should be used when coordinating agents, delegating tasks to specialists, or when "dispatch agents", "which agent", or "multi-agent" are mentioned.

outfitter-dev Bundle

This skill should be used when the user asks to "create a virtual branch", "assign file to branch", "work on multiple features simultaneously", "organize commits after coding", "use but commands", or mentions GitButler, virtual branches, parallel development without checkout, post-hoc commit organization, multi-agent concurrent development, or `--gitbutler`/`--but` flags.

outfitter-dev Bundle

This skill should be used when creating skills, writing SKILL.md files, or when "create skill", "new skill", "validate skill", or "SKILL.md" are mentioned. Covers cross-platform Agent Skills specification.

tachyon-beep Bundle

Use when designing or reviewing security architecture - threat modeling (STRIDE), defense-in-depth controls, secure-by-design patterns, LLM/AI security (OWASP LLM Top 10, MITRE ATLAS), supply-chain security (SLSA/SBOM/Sigstore), compliance mapping (NIST CSF/ISO 27001/PCI-DSS/GDPR), classified-systems MLS, and government authorization (ATO/RMF/FedRAMP).

tachyon-beep Bundle

Use when a model's OUTPUT is a graph or structure rather than a label or scalar - generating neural-architecture candidates, program-synthesis over typed IRs, molecule or circuit graphs, or any typed DAG emitted by a generator. Covers typed graph grammars with node/edge/cost ceilings, deterministic and latent-conditioned generation, best-of-K pools, mutation/recombination over lineages, canonicalisation to normal forms, graph isomorphism and equivalence detection, semantic hashing, and mode-collapse diagnosis. Use for "graph grammar", "DAG generation", "canonicalise a graph", "graph isomorphism", "semantic hash", "duplicate rate after canonicalisation", "constrained decoding" for structured outputs, or "evolutionary search over graphs". Also use to audit a generator/verifier split for the generator grading its own examination.

tachyon-beep Bundle

Use when selecting or comparing neural architectures - routes by data modality (vision / sequence / graph / generative / multimodal) and constraints (dataset size, compute, latency); covers CNNs (ConvNeXt v2 / EfficientNetV2), Transformers + MoE (Mixtral / DeepSeek), SSM/Mamba, modern diffusion (SDXL/FLUX/DiT), multimodal (CLIP/SigLIP/LLaVA), SAM/SAM-2, equivariant GNNs, normalization, attention variants

mgriot Bundle

The ultimate authority for engineering, architecting, and optimizing Agent Skills across any LLM platform. Use for creating new skills from scratch, modifying and improving existing skills, and measuring skill performance. You MUST use this skill whenever the user mentions creating, modifying, or evaluating agent skills, prompts, or specialized workflows. Use it even if the user only asks for "help with instructions" or wants to "test a behavior," as the Master Architect provides the most rigorous framework for intent capture, RED-baseline testing, benchmarking with variance analysis, dynamic context injection, and description optimization for better triggering accuracy.

outfitter-dev Bundle

This skill should be used when creating agents, writing agent frontmatter, configuring subagents, or when "create agent", "agent.md", "subagent", or "Task tool" are mentioned.

outfitter-dev Bundle

This skill should be used when configuring Claude, setting up MCP servers, or when "settings.json", "claude_desktop_config", "MCP server", or "Claude config" are mentioned.

outfitter-dev Bundle

This skill should be used when creating Claude Code skills with Claude-specific features like allowed-tools, context modes (fork/inherit), argument-hint, or model overrides. Triggers on "Claude skill", "allowed-tools", "context fork", "skill arguments".

tachyon-beep Bundle

Use for counterfactual and paired-branch ML experiments - matched-seed comparisons, branch rollouts from a shared snapshot, ablation forests, best-of-K screening. Fires on "is this significant", "how many runs/seeds do I need", paired comparison, counterfactual evaluation, selection bias, winner's curse, best-of-K, abstention calibration, data leakage between splits, pre-registration, Pareto frontier reporting, pseudo-replication. Covers the independent statistical unit (the trajectory, not the branch), cluster-robust inference, paired tests against a zero-anchored no-op control, common random numbers, grouped splits, multiple comparisons and sequential testing, paired power analysis, cost-charged utility, and reliability reporting.

tachyon-beep Bundle

Use when building or auditing a compiler for tensor programs — lowering a graph IR to executable PyTorch, writing torch.fx passes, driving torch.compile/dynamo/inductor/AOTAutograd, selecting kernels, fusing operators, or planning memory. Use when compiled output differs from eager, gradients mismatch after compilation, a numerical tolerance is being argued about, a compilation cache returns a stale or wrong artifact, or model builds must be reproducible. Covers semantic identity vs execution strategy, conformance gates independent of the compiler (reference execution, gradcheck, cross-device and cross-layout agreement), compilation manifests, content-addressed artifacts, compile-time budgets and staleness, and miscompile bisection. Producer-side pack — it transforms IR and emits executables. To read code and produce verdicts use `/static-analysis-engineering`; for run-versus-replay divergence use `/determinism-and-replay`.

outfitter-dev Skill

This skill should be used when dispatching subagents for parallel development, coordinating multi-branch implementations, or when "parallel agents", "orchestrator commits", "subagent filesystem only", "multi-agent git", or "prevent stack corruption" are mentioned. Prevents stack corruption through orchestrator-only git policy.

outfitter-dev Bundle

Transforms external repositories (CLIs, libraries, MCP servers) into Claude Code plugins with skills. Use when "build plugin for", "create skills for CLI", "package as plugin", "repo to plugin", or "turn into plugin" are mentioned.

outfitter-dev Skill

Troubleshoots Outfitter Stack issues including Result handling, MCP problems, CLI output, exit codes, and logging. Use when debugging stack-specific issues, unexpected errors, wrong output modes, or when "debug Result", "MCP not working", "wrong exit code", or "logging issue" are mentioned.

metinduraktr-44 Skill

Routes Cursor plugin skills and MCPs into AdOps LLM mini-agencies by family. Use when the user names a /skill, asks to run all skills, or requests MCP routing for the agency.

outfitter-dev Bundle

Reference for Outfitter Stack patterns including Result types, Handler contract, Error taxonomy, and @outfitter/* package conventions. Use when learning the stack, looking up patterns, understanding packages, or when "Result", "Handler", "error taxonomy", "OutfitterError", "CLI output", "pagination", "MCP server", "MCP tool", "structured logging", "redaction", "test handler", "daemon", "IPC", or "@outfitter/*" are mentioned.

outfitter-dev Bundle

Templates for creating handlers, CLI commands, MCP tools, and daemon services following Outfitter Stack conventions. Use when scaffolding new components, creating handlers, adding commands, or when "create handler", "new command", "add tool", "scaffold", "template", or "daemon service" are mentioned.

tekliner Bundle

Orchestrate parallel CLI agents (Claude Code, Codex, Gemini) for competitive evaluation. Use when user says "run multi-agent", "compare agents", "launch competitive evaluation", "use parallel agents", or complex tasks rated above 7/10 where multiple approaches exist and best solution matters.

ingpoc Skill

Add, inspect, authenticate, or remove MCP server registrations for the current Codex runtime. Use when the operator asks to configure an MCP server or diagnose whether a registered server is available. Do not clone server code into projects, assume every project needs the same servers, or manage app-specific business workflows.

ingpoc Bundle

Design machine-checkable assertions, stable hashes, fixtures, and exit-code contracts for behavior that must be reproducible. Use when an outcome currently depends on subjective agent judgment or when prompts and generated artifacts need versioned integrity checks. Do not use merely to run an existing repository verification gate; that is verify's job.

ingpoc Bundle

Snapshot only resume-relevant tactical context and an explicitly resumable Codex goal to `.session/CURRENT.md`. Triggers: `/save-session`, `save session`, `save progress`, `checkpoint`. Use when the next agent needs a compact, time-bounded handoff with the exact goal, first action, validation, blockers, route, and context to avoid. Do not capture noisy recent files or treat the checkpoint as permanent truth. Use the shell command `save-session` for the deterministic write path.

ingpoc Bundle

Maintain the canonical local agent-engineering knowledge corpus under ~/.codex/knowledge: ingest an accepted source, refresh an existing article from its primary source, repair metadata, reindex search, or rebuild the generated browser. Use only when the operator asks to change or maintain that corpus. Do not use for evaluating whether research should be adopted or for recalling session history.

ingpoc Bundle

Use when closing meaningful sessions or auditing agent-harness inefficiency: capture preventable friction, optionally mine recent prior sessions for recurrence, choose the narrowest durable control, route it to one owner, and apply obvious safe controls immediately.

ingpoc Bundle

Bootstrap or assess a Codex-first day-one readiness baseline for an unfamiliar repo. Use when entering a repo for the first time, making a repo agent-runnable, or needing one canonical readiness report before implementation.

technicalpickles Bundle

Save current work context for later resumption