238 skills · sorted by relevance
Build and evaluate single-agent reinforcement learning experiments, where an agent learns actions from rewards. Use Stable Baselines3 to choose algorithms, create environments, monitor training, and save models with the normalization data needed for consistent evaluation.
Audit an AI agent's completion claims using existing logs, artifacts, approvals, and deployment records. Reconstruct the run and assess each claim separately, distinguishing proven results, partial support, contradictory evidence, and missing proof without rerunning the work.
Set up Antigravity 2.0 Agent Manager alongside Antigravity IDE to coordinate parallel development agents. Assign roles and folder boundaries, review their changes in the editor, and use Git checkpoints to manage shared work.
Evaluate whether a fine-tuned model is ready for release by checking training data, general capability loss, comparisons with the base model, and staged production rollout. Produces an evidence-based promotion or rejection report with one prioritized remedy when rejected.
Design vocational training around a real work task, from preparation and equipment checks to operation, verification, and handover. Build interactive practice with tools, ordered steps, consequences for mistakes, and checkpoints where learners decide whether to proceed, recheck, or stop.
Design a slide-based lecture with sustained teacher narration, connected arguments, and recurring real cases. It organizes ideas into short sequences and places occasional application quizzes at transitions, helping learners understand a subject well enough to explain it.
Provides prompts for assessing and planning .NET upgrades across multiple projects. Helps identify compatibility issues, order migration work, propose code and build pipeline changes, and prepare testing checklists, review guidance, and release documentation.
Convert feature requirements and implementation plans into a test strategy, task checklist, and quality assurance plan for GitHub projects. Define test coverage, priorities, dependencies, estimates, and quality checkpoints using established testing and software quality frameworks.
Guide a development brief through specification, dependent tickets, implementation, review, and verification. Coordinate parallel work while keeping human approval at key decisions and recording agreed vocabulary, business rules, and the reasons behind technical choices.
Ask a more capable agent for a concrete execution plan, including step order, dependencies, and checks. The requesting agent then adapts and carries out the plan, with follow-up questions available to clarify individual steps.
Check training data and machine learning models for malicious changes, including backdoors that cause unexpected behavior on specific inputs. Combine data quality checks, model analysis, and file integrity verification, then document findings and retest after remediation.
Assess language models with over 60 academic benchmark tests using consistent prompts and scoring. Compare models, track performance during training, and organize research results with support for HuggingFace models, vLLM, and APIs.
Coordinates specialist agents in Claude Code to work from product requirements through design, implementation, testing, deployment, and operations. Includes task queues, saved progress, parallel reviews, and recovery workflows for managing extended development work.
Organize PyTorch training code with LightningModule and Trainer. Follow examples for validation, logging, model saving, and distributed training across multiple GPUs, with callbacks that add monitoring or stop training when progress stalls.
Process large email, document, and transcript archives by routing items to AI models according to their importance. Rule-based privacy checks run before model calls, while extracted information becomes linked knowledge pages with progress checkpoints.
Plan database, service, and infrastructure migrations with phased steps, compatibility checks, and rollback procedures for returning to the previous system. Define risks and validation checkpoints to guide transitions with minimal disruption to business operations.
Prepare a long play session with timed observation checkpoints and logs for memory growth, stability, and player fatigue. It creates the testing protocol and analysis template; a human plays the game and records the findings.
Review a development milestone using sprint reports, feature status, bugs, and risks. Summarize completion, quality, and delivery pace, then recommend whether to proceed, proceed with conditions, or stop, with scope choices and action items.
Write tests that demonstrate a pull request’s user-visible changes through before-and-after steps, screenshots, and recorded observations. Extend an existing core journey or create a one-off proof that helps reviewers understand the outcome and access boundaries.
Develop patent application drafts from an invention description, covering prior art searches, claims that define protection scope, specifications, and examiner-style review. Compile documents for China, the United States, or Europe, with human checkpoints throughout.
Improve a compiled paper through two rounds of external model review, source edits, and PDF rebuilding. Check claims against evidence, resolve inconsistencies, and refine figures and layout, with optional checkpoints and restrictions on which edits are allowed.
Assess language models with over 60 academic benchmark tests using consistent prompts and scoring. Compare models, track performance during training, and organize research results with support for HuggingFace models, vLLM, and APIs.
Resume a project by restoring its saved context, locating checkpoints, and identifying unfinished plans. The skill presents the current status and routes work to an appropriate next action based on what it finds.
Coordinate work across an OpenRig agent team with clear assignments, attached context, completion criteria, and selected review checkpoints. Track queue status, identify blocked or idle owners, and resolve handoff gaps while leaving implementation with the assigned agents.