Loading...
Loading...
Found 172 Skills
Use when "training LLM", "finetuning", "RLHF", "distributed training", "DeepSpeed", "Accelerate", "PyTorch Lightning", "Ray Train", "TRL", "Unsloth", "LoRA training", "flash attention", "gradient checkpointing"
FORGE Autopilot — Intelligent autonomous mode. FORGE analyzes the project state, automatically decides the next action, and orchestrates all agents until completion. Configurable checkpoints for human review. Usage: /forge-auto or /forge-auto "specific objective"
Evidence-based memory optimization from real usage patterns. Analyzes recall performance, identifies bottlenecks, suggests consolidation/pruning/enrichment, and tracks improvement over time via checkpoint Q&A.
Conduct simulated user research with AI personas. Triggers when the user says 'do user research', 'run user research', 'simulate user interviews', or '/user-research'. Three phases: free growth → pain extraction → product collision, with four quality validation checkpoints. Supports single or multi-concept testing.
Run the full spec-driven workflow automatically. Proposes, implements, verifies, reviews, and archives a change with one mandatory proposal checkpoint plus any extra confirmations required by blocking conditions.
Use when breaking work into discrete steps, tracking progress through multi-step implementations, or managing implementation task lists. Triggers when an approved plan needs to be converted into tracked tasks, when progress reporting is needed during execution, or when checkpoint reviews are required between task batches.
Build a pre-implementation harness for ambiguous or risky coding tasks by grounding the request in the repository, producing a structured impact map, surfacing ambiguities and risks, defining scope boundaries, and creating a validation-ready implementation contract before any code changes are made. Use when a task is broad, underspecified, cross-cutting, or likely to drift without an explicit planning checkpoint.
Design state machines, orchestration workflows, saga patterns, and resilience strategies for distributed systems, AI agents, and complex async processes. Use when asking for a workflow, state machine, orchestration design, saga, HITL checkpoint, or process resilience strategy.
Use when inspecting, cleaning, understanding, reproducing, or auditing academic research code repositories, especially when README commands, datasets, checkpoints, experiments, or paper claims need verification.
End-to-end pipeline from unlabeled ml_app traces to a bootstrapped evaluator suite. Runs trace classification → root cause analysis → eval bootstrap in sequence with user checkpoints. Use when user says "run the eval pipeline", "go from traces to evals", "bootstrap evals end to end", "classify then RCA then bootstrap", "build an eval set from scratch", or wants a guided walkthrough from production data to evaluator code.
Brev instance operating guidance for NeMo-RL agents working in /home/ubuntu/RL with limited workspace disk, a larger /ephemeral volume, and optional /home/ubuntu/RL/.env secrets. Use when running auto-research campaigns, experiments, training jobs, model or dataset downloads, shared cache-heavy commands, log-producing runs, checkpoint generation, W&B or Hugging Face authenticated workflows, or any workflow that may create large files on Brev.
Build a personalized learning roadmap with milestones and practice checkpoints