Claude Skills by stanfish06

Create LangChain agents with create_agent, define tools, and use middleware for human-in-the-loop and error handling.

INVOKE THIS SKILL when you need human-in-the-loop approval, custom middleware, or structured output. Covers HumanInTheLoopMiddleware for human approval of dangerous tool calls, creating custom middleware with hooks, Command resume patterns, and structured output with Pydantic/Zod.

Scaffold a minimal local LangChain agent in Python by following the official quickstart. Use when the user wants to quickly build or try a LangChain agent locally.

INVOKE THIS SKILL when building ANY retrieval-augmented generation (RAG) system. Covers document loaders, RecursiveCharacterTextSplitter, embeddings (OpenAI), and vector stores (Chroma, FAISS, Pinecone).

Scaffold a minimal local LangChain agent in TypeScript by following the official quickstart. Use when the user wants to quickly build or try a LangChain agent locally.

Interact with Langfuse and access its documentation: tracing, monitoring, creating datasets, running experiments, and evaluating AI applications. Use when needing to (1) query or modify Langfuse data, (2) look up Langfuse documentation, concepts, integration guides, a feature or SDK usage, or (3) do any AI engineering task (AI observability, prompt engineering/management, evaluation and evaluator management, experimentation, dataset management, evaluation-driven CI/CD, feedback collection). I...

INVOKE THIS SKILL when using the langgraph CLI to scaffold, develop, build, or deploy LangGraph applications. Covers langgraph new, dev, build, up, deploy, and langgraph.json configuration.

INVOKE THIS SKILL when writing ANY LangGraph code. Covers StateGraph, state schemas, nodes, edges, Command, Send, invoke, streaming, and error handling.

INVOKE THIS SKILL when implementing human-in-the-loop patterns, pausing for approval, or handling errors in LangGraph. Covers interrupt(), Command(resume=...), approval/validation workflows, and the 4-tier error handling strategy.

INVOKE THIS SKILL when your LangGraph needs to persist state, remember conversations, travel through history, or configure subgraph checkpointer scoping. Covers checkpointers, thread_id, time travel, Store, and subgraph persistence modes.

Scaffold a minimal local LangGraph agent in Python by following the official quickstart. Use when the user wants to quickly build or try a LangGraph agent locally.

Scaffold a minimal local LangGraph agent in TypeScript by following the official quickstart. Use when the user wants to quickly build or try a LangGraph agent locally.

LangGraph is a low-level orchestration framework for building stateful LLM agents and workflows as explicit graphs — typed state with reducers, nodes/edges/conditional routing, checkpointer-backed persistence and thread memory, human-in-the-loop interrupts, and streaming. Use when building an agent or multi-step LLM workflow that needs explicit control flow (branching, loops, parallel fan-out), durable/resumable execution, approval gates, or multi-agent handoffs — or when debugging StateGraph...

Iteratively inspect traces, interview the user, and create LangSmith online evaluators one at a time. Use specifically for creating online evaluators for use within LangSmith -- use "eval-engineering" for Harbor-style online evaluations.

Expert-thinking profile for Laser Physicist (laser source R&D / ultrafast & CPA / nonlinear frequency conversion / beam metrology / laser safety (ISO 11146, IEC 60825)): Expert profile for laser physicist — see AGENTS.md for field-specific methods and failure modes.

Latch platform for bioinformatics workflows. Build pipelines with Latch SDK, @workflow/@task decorators, deploy serverless workflows, LatchFile/LatchDir, Nextflow/Snakemake integration.

Create professional research posters in LaTeX using beamerposter, tikzposter, or baposter. This is the default poster skill for conference presentations, academic posters, and scientific communication. Includes layout design, color schemes, multi-column formats, figure integration, and poster-specific best practices. For editable PowerPoint output, use pptx-posters (HTML/CSS-based) or poster-presentation (python-pptx-based) instead.

Launch Readiness Orchestrator skill for Datasite deal rooms. Use this skill whenever a deal team wants a single pre-go-live readiness check across their data room — combining gap analysis, document quality audit, and risk review into one consolidated "is the room ready?" view. Triggers include: "are we ready to go live", "launch readiness check", "pre-launch audit", "data room readiness", "can we launch", "is the data room ready", "run a full readiness check", "go-live checklist", "pre-launch...

Compute pairwise r² between a lead variant and every variant in a window using the 1000 Genomes Phase 3 GRCh38 reference panel, ancestry-stratified. Use when an agent needs LD coloring for a regional plot or LD pruning around a candidate causal variant. Single client (on-demand region fetch from EBI 1000G FTP); no multi-GB cold-start.

Expert-thinking profile for Life-Cycle Assessment Analyst (attributional/consequential LCA / inventory & impact assessment / allocation & uncertainty / EPD critical review (ISO 14040/14044, EN 15804)): Reasons from functional unit, attributional-versus- consequential framing, and ISO 14044 allocation hierarchy through openLCA, SimaPro, Brightway2, ecoinvent, and LCIA methods like TRACI and EF 3.0 while treating allocation-driven ranking flips, biogenic-versus-fossil carbon mistagging, cut-off...

Expert-thinking profile for Limnologist (field / observational / physical limnology / lake ecology / paleolimnology): Reasons from stratification, Schmidt stability, and nutrient–light coupling; profiles with CTD/EXO and Carlson TSI components; models with rLakeAnalyzer and GLM while treating internal P loading, sensor fouling, and spatial pseudoreplication as first-class failure modes.

Implement and review macOS SwiftUI Liquid Glass UI. Use when adopting system glass, removing conflicting custom chrome, or building glass surfaces.

'Search PubMed and bioRxiv for bioinformatics literature, synthesise results into a structured report, and build

Local document and PDF parsing with spatial text and bounding boxes. Use for extracting text from PDFs, DOCX, Office files, and images; OCR on scans; layout-preserved JSON for RAG; batch-ingesting paper folders; or page screenshots for multimodal agents — even when the user does not name liteparse. Prefer over MarkItDown when you need bboxes, fast local parsing, or PNG page renders; prefer over the pdf skill for merge/split/forms.

Conduct comprehensive, systematic literature reviews using multiple academic databases (PubMed, arXiv, bioRxiv, Semantic Scholar, etc.). Use when conducting systematic literature reviews, meta-analyses, research synthesis, or comprehensive literature searches across biomedical, scientific, and technical domains. Creates professionally formatted markdown documents and PDFs with verified citations in multiple citation styles (APA, Nature, Vancouver, etc.).

Writes public-facing litigation updates — blog posts, client alerts, LinkedIn/X posts — on a federal case or legal development. Use to \"write a blog post about…,\" \"draft a client alert on…\"

Expert guidance for LlamaIndex development including RAG applications, vector stores, document processing, query engines, and building production AI applications.

LLM and agent security red teaming with agentic-actions-auditor, supply-chain-risk-auditor, semgrep, codeql, and sarif-parsing. Use for prompt injection, tool misuse, data exfiltration, excessive agency, insecure output handling, supply-chain risk, retrieval poisoning, and OWASP LLM Top 10 style threat modeling.

LLM and agent observability, tracing, and evaluation workflows with langfuse, phoenix-cli, and phoenix-evals. Use when instrumenting Langfuse, Phoenix, OpenTelemetry GenAI traces, eval datasets, prompt experiments, latency/cost debugging, trace scoring, or regression testing agent behavior.

Use for authorized security assessment of LLM applications and AI agents, including prompt injection, tool abuse, RAG exposure, memory poisoning, and model supply-chain risks.

Render a 4-panel regional LocusCompare diagnostic for one (lead variant, exposure study, outcome study) tuple - overlays GWAS Manhattan, QTL Manhattan, GENCODE gene track, and cross-trait scatter colored by LD r². Use when an agent needs visual confirmation that two GWAS / QTL signals share the same causal variant (the Liu 2019 LocusCompare convention). Inputs: lead variant + two pre-fetched harmonised sumstats slices (or eQTL Catalogue / GWAS Catalog identifiers for bundled fetch). Output: P...

Add Pydantic Logfire observability to applications and send as much useful telemetry as possible. Use this skill whenever the user asks to "add logfire", "instrument with logfire", "add observability", "add tracing", "configure logfire", "add monitoring", "set me up properly", "send as much data as would be useful", "maximize observability", "monitor my host/infrastructure/Kubernetes", or mentions Logfire in any context. Supports Python, JavaScript/TypeScript, and Rust, plus host/infrastructu...

Query and analyze Logfire telemetry data — traces, logs, spans, metrics, summaries, and SQL results. Use this skill when the user asks to "query logfire", "search traces", "find logs", "query data", "search spans", "look up errors in logfire", "get metrics from logfire", "analyze telemetry", "summarize errors", "find root cause", or add Logfire querying capabilities to code. Do not use this skill for direct Logfire UI, browser, live-view, Explore-page, or link-opening requests; use logfire-ui...

Open or return Logfire project pages, live views, trace links, and Explore pages in the Codex browser without querying telemetry first. Use this skill when the user asks to "open in Logfire", "show in the live view", "open Explore", "open the UI", "show in Codex", "use the browser", "give me a link", or asks for a Logfire GUI/browser/live-view presentation of a project, time range, service, span, trace, log, or filter. If "show" or "view" wording is ambiguous, ask whether the user wants a UI ...

Expert-thinking profile for Logician (proof theory / model theory / computability / set theory / formal verification (Lean, Coq, Z3)): Expert profile for logician — see AGENTS.md for field-specific methods and failure modes.

Expert evaluator for Grafana Loki label strategy. Audits, designs, and improves label schemas using cardinality scoring, access-pattern alignment, static vs. dynamic label rules, and consistency checks. Use when the user asks to evaluate, audit, design, or improve a Loki label strategy — or asks why their Loki queries are slow.

Grill me about specs for the workflows I want to build, within this workspace.

Expert-thinking profile for Low-Temperature Physicist (experimental / cryogenic / condensed-matter & quantum transport): Reasons from kT budgets, He-3/He-4 dilution refrigeration, and BCS/GL superconductivity; measures Tc, QHE, and Landauer conductance with lock-in/SQUID workflows while treating wiring heat loads, Kapitza resistance, flux trapping, TLS dielectric loss, and sample-vs-MXC thermometer mismatch as first-class failure modes.

Expert-thinking profile for Machine Learning Engineer (production ML / MLOps / feature stores & serving): Reasons from feature-store point-in-time joins (Feast/Tecton), Airflow/Kubeflow training pipelines, MLflow registry, Triton/TorchServe/BentoML serving, Evidently/WhyLabs drift and PSI, shadow/canary/A/B rollouts, inference SLAs, and reproducible training hashes while treating train–serve skew, label leakage, and...

Expert-thinking profile for Machine Learning Researcher (computational / empirical & theoretical machine learning research): Reasons from population risk, double descent, and inductive bias; enforces sacred test sets, hierarchical ablations, nested CV, and HELM/Dynabench-aware benchmarking; reports with NeurIPS and Pineau reproducibility checklists while treating leakage, meta-overfitting, benchmark contamination, Goodhart gaming, and seed...

Use for authorized macOS and Mach-O reverse engineering including codesign, Objective-C/Swift recovery, endpoint security surfaces, and Apple platform malware analysis.

Use when the user mentions MagicPath, designs, UI components, themes, canvas selections, or repo-to-canvas UI work; run magicpath-ai to search, inspect, install, or author components.

Use when analyzing suspected malware through static, dynamic, and behavioral techniques, including IOC extraction, YARA or Sigma rules, sandboxing, and anti-analysis behavior.

Expert-thinking profile for Mammalogist (field / observational / museum systematics / telemetry): Reasons from mammalian life history and detectability-limited sampling through ASM MDD taxonomy, Sherman/camera-trap/SCR survey design, occupancy and SECR models, bat acoustic validation, and museum voucher discipline while treating trap heterogeneity, camera autocorrelation, closure violation, and WNS decontamination...

INVOKE THIS SKILL when building, testing, or deploying Managed Deep Agents in LangSmith with the mda CLI. Walks a user through their first agent end to end — interviewing them about what they want to build, mapping it onto what MDA can actually do, then scaffolding and deploying it. Covers the file-based project layout; define_deep_agent / defineDeepAgent; instructions, skills, memory, identity, tools, middleware, sandboxes, schedules, channels, and evals; mda init/build/dev/deploy/logs/delet...

Expert-thinking profile for Manufacturing Engineer (process planning / CNC-CAM / GD&T / quality launch (APQP/PPAP, MSA/SPC) / multi-process (machining, welding, casting)): Reasons from process physics, capability, and cost through Shercliff-Lovatt process selection, ASME Y14.5 GD&T, CAM simulation, and AIAG APQP/PPAP with MSA-gated SPC capability, treating high %GRR masquerading as variation, false Cpk on unstable or short runs, datum-scheme mismatch, and uncontrolled ECN tweaks as...

Reactive Python notebooks stored as pure .py files — cells form a dependency DAG, so changing one cell automatically reruns its dependents (or marks them stale under the lazy runtime) and hidden state cannot exist. Covers the notebook file format, marimo edit/run/export CLI, mo.ui interactive elements, SQL cells, running notebooks as apps or scripts, and sandboxed notebooks with PEP 723 inline dependencies. Use when creating or editing marimo notebooks, building interactive data apps or dashb...

Expert-thinking profile for Marine Biologist (field / shipboard / biological oceanography / eDNA & observing systems): Reasons from water-mass stratification, CTD–Niskin and CalCOFI-style net tows, BRUV, and MiFish/COI eDNA through OBIS/WoRMS/GBIF and ARGO/BGC-Argo; treats mesopelagic DVM, hypoxia/Ω_aragonite constraints, fluorometer quenching, BRUV MaxN bias, and transect pseudoreplication as first-class failure modes.

Expert-thinking profile for Marine Engineer (shipboard / design / machinery systems / class compliance): Reasons from propulsion thermodynamics, shaft BPF/torsional barred speeds, central LT/HT cooling, class machinery surveys, and ISO 15016:2025 sea trials while treating cat fines liner wear, scavenge fire, purifier mis-set, blackout PMS logic, and tropical SW fouling as first-class failure modes.

Expert-thinking profile for Marine Geologist (ship/AUV survey / seismic + multibeam / IODP coring + stratigraphy / marine geohazards + seafloor resources): Reasons from stratigraphy, sedimentary processes, geophysical facies, and age control through multibeam bathymetry, 2D/3D seismic, piston/IODP cores tied via synthetic seismograms, and CSF-A age models while treating bad-SVP false scarps, BSRs mimicking free gas, gas-charged push-down faking structural offset, and...