Menu
SkillsMP has collected 122 skills from TheAdaply/re-forge. Open a skill to review its source and details.
Showing 40 of 122 collected skills.
Correlation-to-causation discipline for any "does X cause Y", "what is the effect of X", or "should we do X" question. Use when a decision rests on a causal claim drawn from observational data, an A/B test, an event study, or a regression with controls —…
Provides guidance for interpreting and manipulating neural network internals using nnsight with optional NDIF remote execution. Use when needing to run interpretability experiments on massive models (70B+) without local GPU resources, or when working with any…
Use this skill for long-horizon agentic coding where a feature spans many turns and the agent drifts: it freezes a re-readable contract before code. Writes EARS-style requirements -> design -> task breakdown -> verification evidence, one feature directory per…
Build a backtest that will not lie to you — the engineering, not the statistics. Use when implementing, reviewing, or debugging a backtest/simulation of a trading or sequential-decision strategy: choosing event-driven vs vectorized, auditing…
Run a Bayesian analysis as an ordered workflow, not a single fit: state and justify priors, prior-predictive-check before fitting, run the full Gelman-et-al workflow loop, gate on convergence diagnostics (R-hat, ESS, divergences), posterior-predictive-check…
Data-quality engineering that runs BEFORE any analysis or model — even on one CSV. Use when ingesting, joining, transforming, caching, or reconciling data, or whenever a downstream number depends on a table being correct, fresh, and reproducible. Enforces…
Structured red-teaming for any analysis, plan, or conclusion before it ships. Use when about to commit to a decision, recommend an approach, accept a result, or write a conclusion that others will act on — especially when the room agrees, the number looks…
Risk thinking for portfolios and any decision under uncertainty. Use when sizing a bet, allocating capital, deploying a strategy, or making any irreversible call where a bad outcome can compound: ask the ruin question first (the bet size from which you cannot…
Enforce hypothesis-testing discipline before you trust any p-value, win, or significance claim. Use when running an A/B test, comparing models or strategies, screening features, declaring a result 'significant', or deciding whether a null result means…
Forecasting discipline for any time-ordered data — sensor logs, demand, prices, request rates, KPIs. Use when asked to forecast, predict the next value(s), model a series with a time index, test stationarity, decompose trend/seasonality, validate a forecast,…
Run a tabular/ML competition end to end the way a grandmaster does: metric-first scoping, leakage forensics, a validation scheme that mirrors the test split before any model, a dumb baseline, one-change-per-experiment iteration, winner-derived…
Procedures for competing in LLM-era ML competitions — NLP/text classification, LLM fine-tuning tracks, and prompt/agent contests. Use when the task is a text or LLM competition graded by a hidden metric under latency/memory caps: rebuild the metric locally…
Interrogate any dataset the way a quantitative analyst does before trusting a single number from it — even a basic CSV. Use when asked to analyze, explore, model, backtest, or draw conclusions from data with a time dimension, a predictive target, many…
Run re-forge teams as deterministic ultracode Workflow scripts in Claude Code: fan out one auditor per file/batch with adversarial refuters, gate behind a strict-schema synthesis agent. Use when the work is enumerable (audit N files, refute M claims) rather…
Make a software project score the highest possible rating from an LLM evaluator (Claude Opus, GPT, code-review bots, preference models) AND from expert human reviewers — by building genuine, machine-verifiable quality and legibility, never by gaming judge…
Write publication-ready ML/AI/Systems papers for NeurIPS, ICML, ICLR, ACL, AAAI, COLM, OSDI, NSDI, ASPLOS, SOSP. Use when drafting papers from research repos, structuring arguments, verifying citations, or preparing camera-ready submissions. Includes LaTeX…
Inventory the operator's workforce (agents, skills, plugins, installed MCP servers, recent research team SYNTHESIS files) and produce a ranked capability gap list. Diffs against VoltAgent's 10-category specialist taxonomy and wshobson/agents to find missing…
Move a tested skill draft from ~/.claude/forge/drafts/ to ~/.claude/skills/<name>/ for personal use or to ~/.claude/forge/outputs/<plugin>/ for plugin packaging. Updates the Forge's MEMORY.md catalog with a bullet entry (authored_at, helpful_count=0,…
Query external sources (MCP Registry, anthropics/skills, installed marketplaces) for candidate implementations of a specified capability. Applies the 6-rule trust heuristic to each candidate and returns ranked recommendations. Use when forge-gap has…
Run the skill-creator eval loop against a forge-draft output to grade whether the drafted skill is promotable. Invokes the grader and analyzer agents plus the aggregate-benchmark script from the official skill-creator plugin and returns a pass/fail verdict +…
Orchestrates end-to-end autonomous AI research projects using a two-loop architecture. The inner loop runs rapid experiment iterations with clear optimization targets. The outer loop synthesizes results, identifies patterns, and steers research direction.…
Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace…
PyTorch library for audio generation including text-to-music (MusicGen) and text-to-sound (AudioGen). Use when you need to generate music from text descriptions, create sound effects, or perform melody-conditioned music generation.
Autonomous AI agent platform for building and deploying continuous agents. Use when creating visual workflow agents, deploying persistent autonomous agents, or building complex multi-step AI automation systems.
Expert guidance for fine-tuning LLMs with Axolotl - YAML configs, 100+ models, LoRA/QLoRA, DPO/KTO/ORPO/GRPO, multimodal support
Quantizes LLMs to 8-bit or 4-bit for 50-75% memory reduction with minimal accuracy loss. Use when GPU memory is limited, need to fit larger models, or want faster inference. Supports INT8, NF4, FP4 formats, QLoRA training, and 8-bit optimizers. Works with…
Vision-language pre-training framework bridging frozen image encoders and LLMs. Use when you need image captioning, visual question answering, image-text retrieval, or multimodal chat with state-of-the-art zero-shot performance.
Guides researchers through structured ideation frameworks to discover high-impact research directions. Use when exploring new problem spaces, pivoting between projects, or seeking novel angles on existing work.
Open-source embedding database for AI applications. Store embeddings and metadata, perform vector and full-text search, filter by metadata. Simple 4-function API. Scales from notebooks to production clusters. Use for semantic search, RAG applications, or…
OpenAI's model connecting vision and language. Enables zero-shot image classification, image-text matching, and cross-modal retrieval. Trained on 400M image-text pairs. Use for image search, content moderation, or vision-language tasks without fine-tuning.…
Anthropic's method for training harmless AI through self-improvement. Two-phase approach - supervised learning with self-critique/revision, then RLAIF (RL from AI Feedback). Use for safety alignment, reducing harmful outputs without human labels. Powers…
Evaluates NVIDIA Cosmos Policy on LIBERO and RoboCasa simulation environments. Use when setting up cosmos-policy for robot manipulation evaluation, running headless GPU evaluations with EGL rendering, or profiling inference latency on cluster or local GPU…
Applies cognitive science frameworks for creative thinking to CS and AI research ideation. Use when seeking genuinely novel research directions by leveraging combinatorial creativity, analogical reasoning, constraint manipulation, and other empirically…
Multi-agent orchestration framework for autonomous AI collaboration. Use when building teams of specialized agents working together on complex tasks, when you need role-based agent collaboration with memory, or for production workflows requiring…
Expert guidance for distributed training with DeepSpeed - ZeRO optimization stages, pipeline parallelism, FP16/BF16/FP8, 1-bit Adam, sparse attention
Build complex AI systems with declarative programming, optimize prompts automatically, create modular RAG systems and agents with DSPy - Stanford NLP's framework for systematic LM programming
Facebook's library for efficient similarity search and clustering of dense vectors. Supports billions of vectors, GPU acceleration, and various index types (Flat, IVF, HNSW). Use for fast k-NN search, large-scale vector retrieval, or when you need pure…
Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training/running transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference. Supports…
GGUF format and llama.cpp quantization for efficient CPU/GPU inference. Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization from 2-8 bit without GPU requirements.
Post-training 4-bit quantization for LLMs with minimal accuracy loss. Use for deploying large models (70B, 405B) on consumer GPUs, when you need 4× memory reduction with <2% perplexity degradation, or for faster inference (3-4× speedup) vs FP16. Integrates…