데이터·AI 스킬

Manage a production-critical Marin training run from launch checks through monitoring, recovery, handoff, and completion. Keep commands, code versions, checkpoints, saved training states, and final outputs tied to a durable run record.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Launch a production training run with verified settings
• Resume interrupted training from a complete checkpoint
• Monitor training progress and investigate failures

Define and launch Marin SkyRL reinforcement learning experiments on Iris. It guides configuration checks, model and data dependencies, small trial runs, and progress verification while keeping artifact versions and runtime revisions explicit.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Define a reinforcement learning experiment
• Check resource allocation and role configuration before launch
• Verify training with a small trial run

Register Hugging Face datasets for Marin by inspecting their structure without downloading the full dataset. Identify text fields, configurations, and data splits, then add a dataset module and validate the mapping with a sample.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Registering a Hugging Face dataset for Marin
• Inspecting text fields without a full dataset download
• Checking configurations and data splits

Intersect polygon layers, which describe geographic areas, and measure their overlap after checking coordinate systems and geometry. Produce GeoJSON results, an area summary, and records of inputs and settings for repeatable analysis.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Measure overlap between two geographic area layers
• Check coordinate systems and polygon validity before analysis
• Save intersection results as GeoJSON

Create a local run efficiency report from supplied timings and context limits, the amount of information available to the agent. Include recorded Hermes token usage and costs when available, and clearly identify missing measurements.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Reviewing how a run spent its time
• Checking context use against its budget
• Examining supplied tool execution durations

Guide the addition of a new model generation to oh-my-hermes. The workflow checks model recognition, official specifications, routing rules that select models for tasks, documented pricing, and live access, then validates changes with tests and comparative benchmarks.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Add a new model generation to oh-my-hermes
• Check API model IDs and dated snapshot recognition
• Update model selection chains and related documentation

Plan whether to fine-tune a model on your own data, choose a suitable training method, and define evaluation criteria. Compare candidates with the original model before deciding whether to adopt a trained version.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Deciding whether support replies need model training
• Choosing a training method for available data
• Planning separate training and evaluation datasets

Submit text and independent yes-or-no, multiple-choice, or rating questions to Jev using your own key. Receive probabilities rather than prose, along with the served model, request status, and cost information, including cases where no answer is returned.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Check the probability that text meets a stated criterion
• Compare probabilities across predefined choices
• Evaluate text using ordered rating levels

Review external agent prompt files before adapting them for Hermes. The skill inventories sources, checks file formats and argument placeholders, identifies command name conflicts, and records trust concerns so conversion candidates and unresolved issues are clear before implementation.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Review Claude Code prompt files before moving them to Hermes
• Check Codex prompt formats for compatibility
• Plan mappings for arguments such as $ARGUMENTS

Prepare an implementation handoff for an LLM feature, covering model calls, validated output formats, versioned prompts, document retrieval, and evaluation datasets so future prompt or model changes can be compared against a recorded baseline.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Planning an invoice field extraction feature
• Designing document retrieval for an LLM application
• Preparing evaluations for prompt or model changes

Inspect Hermes model settings, confirm active providers, and prepare model assignments for different roles or task categories. Preview configuration changes for approval, guide provider login, and verify the resulting settings after applying approved changes.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Configuring Hermes models for different roles
• Connecting a model provider or switching accounts
• Adjusting model priorities by task category

Guide the adoption of a new model generation or changed service specifications. Check model recognition, research official documentation, draft behavior-specific instructions, plan routing and pricing updates, and define baseline comparisons to measure the adjustments.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Checking recognition of newly released model IDs
• Reviewing changed reasoning and tool-use specifications
• Drafting instructions for model-specific behavior

Plan how to preserve essential information during long agent tasks within limits on tokens, the units used to measure model input and output. Define budgets, summarization checkpoints, and recovery steps while keeping the full objective intact.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Planning information retention for long research or coding tasks
• Preparing essential context for a new session
• Scheduling summaries and checking for lost information

Scope analysis of supplied CSV, JSON, tables, logs, or other data by defining sources, records, questions, and methods. Plan checks for anomalies and trends, and tie reported findings to observed data while requiring separate evidence for causal claims.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Plan CSV anomaly analysis by segment
• Define the scope of trend analysis in tables
• Organize log pattern and error analysis

Consult a configured external AI for a second opinion on a plan or answer. Prepare the question and context, summarize its critique, and explain which suggestions to accept or reject before deciding what to do next.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Get Claude's critique of a plan before implementation
• Ask Gemini for a second opinion on an answer
• Separate accepted and rejected advisor suggestions

Plan repairs and changes for batch or streaming data pipelines. Define writes that remain consistent when repeated, limit historical reprocessing, map downstream dependencies, and specify quality checks before data reaches its readers.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Planning repairs for duplicate events in an ETL job
• Defining a bounded backfill for missing data
• Assessing downstream effects of schema changes

Work through optimization problems using gradients, which indicate how a function changes. The skill guides method selection, step size choices, and convergence checks, with SciPy examples for conjugate gradient and BFGS methods.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Choosing a method for an optimization problem
• Adjusting gradient descent step sizes
• Investigating oscillation or slow convergence

Choose methods for interpolation, estimating values between known data points, based on spacing, smoothness, and noise. Provides SciPy implementation examples, polynomial calculations, and checks for boundary oscillations, prediction error, and the shape of resulting curves.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Estimating values between known data points
• Choosing between polynomial and spline interpolation
• Considering fitting methods for noisy data

Use Shapely to create points, lines, and polygons from coordinates or text descriptions of shapes. Calculate areas and distances, test spatial relationships, combine or transform shapes, and check or repair invalid geometry, with results returned as JSON.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Check whether a point lies inside a polygon
• Calculate shape areas, perimeters, and distances
• Combine shapes or extract overlapping regions

Review Claude Code sessions using Braintrust traces, records of agent activity. Summarize tool, agent, and skill usage, inspect repeated actions, replay sessions, and track token consumption to understand failures and recurring work patterns.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Reviewing a completed Claude Code session
• Investigating failures and repeated tool calls
• Checking weekly agent and skill usage

Choose LLM settings for theorem proving, code generation, and creative exploration. It provides examples for output limits and response variety, plus proof planning prompts and a method for generating multiple proof attempts and selecting the best candidate.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Choosing LLM settings for theorem proving
• Adjusting output settings for code generation
• Increasing response variety for creative exploration

Write clearer instructions for Agentica and REPL agents by specifying roles, file operations, and output formats. Includes templates for dividing research, coordinating specialists, reviewing solutions, and passing results between agents through files.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Clarifying instructions for reading and editing files
• Dividing research among multiple agents
• Passing findings and plans between agents

Set up Agentica patterns that coordinate multiple AI agents. Choose different perspectives for research, a coordinator with specialists for implementation, repeated generation and critique for refinement, or a panel of agents for majority-based validation.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Research a topic from multiple perspectives
• Coordinate planning, implementation, and review
• Refine solutions through generation and critique

Analyze CSV files with Python and get concise numerical summaries. The skill checks column structure before calculating requested aggregates, includes units when available, and explains assumptions when data is missing or incorrectly formatted.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

• Quick summaries of CSV data
• Calculating requested aggregates
• Reporting assumptions about missing or malformed data