Menu
Skills dans ce dépôt
SkillsMP a collecté 184 skills depuis jluo41/Tools. Ouvrez un skill pour examiner sa source et ses détails.
jluo41/ToolsAffichage de 40 skills collectés sur 184.
Use when deeply analyzing a single paper and producing structured notes on claims, methods, figures, evaluation, strengths, limitations, and related work.
Langue du texte source : anglais
Search published venue papers (IEEE, ACM, Springer, etc.) via Semantic Scholar API. Complements /arxiv (preprints) with citation counts, venue metadata, and TLDR. Use when user says "search semantic scholar", "find IEEE papers", "find journal papers", "venue…
Langue du texte source : anglais
Use when conducting literature reviews, summarizing papers, comparing methodologies, identifying research gaps, or supporting scholarly writing across disciplines.
Langue du texte source : anglais
Communications-domain literature review with Claude-style knowledge-base-first retrieval. Use when the task is about communications, wireless, networking, satellite/NTN, Wi-Fi, cellular, transport protocols, congestion control, routing, scheduling, MAC/PHY,…
Langue du texte source : anglais
Review type specialist for the discovery layer: analyze across sources — judge a claim (prior_art_check / counterevidence -> verdict.md) or map a field (landscape_review / benchmark_landscape -> landscape.md). Dispatches research-lit / comm-lit-review /…
Langue du texte source : anglais
Search and analyze research papers, find related work, summarize key ideas. Use when user says "find papers", "related work", "literature review", "what does this paper say", or needs to understand academic papers.
Langue du texte source : anglais
Idea type specialist for the discovery layer: the ideation loop — generate + rank candidate claims (idea_generation -> ideas.md) and evaluate their novelty (novelty_check -> verdict.md). Dispatches idea-creator and novelty-check. Trigger: generate ideas,…
Langue du texte source : anglais
Generate and rank research ideas given a broad direction. Use when user says "找idea", "brainstorm ideas", "generate research ideas", "what can we work on", or wants to explore a research area for publishable directions.
Langue du texte source : anglais
Verify research idea novelty against recent literature. Use when user says "查新", "novelty check", "有没有人做过", "check novelty", or wants to verify a research idea is novel before implementing.
Langue du texte source : anglais
External-evidence layer, and one of the two EXECUTORS (task is the other — same shape, same rules). One research topic = one discovery-folder running Plan -> Build(opt) -> Execute -> Report, typed Search | Review | Idea; buckets 1_search/2_review/3_idea are…
Langue du texte source : anglais
Read-only specialist for haipipe-project. Reviews project structure, summarizes, generates inventory, prints overview. Touches no files. Called by /haipipe-project orchestrator. Direct invocation works for project audit work.
Langue du texte source : anglais
Modify specialist for haipipe-project. Reorganizes files to fix structural violations: flatten flat tasks into groups, rename mis-numbered task folders, align Track A code with Track B examples, fix broken paired-example references. Called by /haipipe-project…
Langue du texte source : anglais
Stage 4 (AIData) specialist: builds/runs/reviews TfmFn / SplitFn, inspects 4-AIDataStore, loads AIData-layer assets + tensors, merges multi-partition CaseSets via streaming HF Dataset. Called by /haipipe-data; direct invocation works stage-scoped.
Langue du texte source : anglais
Stage 3 (Case) specialist: builds/runs/reviews TriggerFn / CaseFn, inspects 3-CaseStore, loads case-layer assets, runs multi-partition in parallel (embarrassingly parallel). Called by /haipipe-data; direct invocation works stage-scoped.
Langue du texte source : anglais
External-reference specialist: builds/runs/reviews ExternalFn (e_build_external_*.py), inspects ExternalStore, loads dimension + engagement assets, previews joins into Source/Record sets. Called by /haipipe-data (recommended entry); direct invocation works…
Langue du texte source : anglais
Stage 0' (raw cohort) specialist: builds a business-readable picture of how one data point in a raw cohort extract is generated, BEFORE it enters Stage 1 (Source). Trigger: raw, rawstore, 0-rawstore, raw cohort, data point generation, datapoint timeline,…
Langue du texte source : anglais
Stage 2 (Record) specialist: builds/runs/reviews HumanFn / RecordFn, inspects 2-RecStore, loads record-layer assets, supports multi-partition via patient_ids predicate pushdown. Called by /haipipe-data; direct invocation works stage-scoped.
Langue du texte source : anglais
Cross-stage transport specialist: pushes/pulls cohort assets between local _WorkSpace and the configured remote (S3 / GCS / Databricks / Google Drive) via hai-remote-sync; never destructive. Trigger: push, pull, sync, remote, S3, upload, download, fetch,…
Langue du texte source : anglais
Run any Stage 1-4 data pipeline work: parses intent (stage + function) and dispatches to the right specialist (source/record/case/aidata, plus raw/external/remote). Use for SourceFn/RecordFn/CaseFn/TfmFn/SplitFn builds, runs, dashboards, reviews, or any…
Langue du texte source : anglais
Stage 1 (Source) specialist: builds/runs/reviews SourceFn, inspects 1-SourceStore, loads source-layer typed frames. Called by /haipipe-data (recommended entry); direct invocation works stage-scoped.
Langue du texte source : anglais
data-pipeline task-folder specialist: scaffolds AND executes {NN}_<name>/ task-folders that run Stage 1-4 builders (Source/Record/Case/AIData) with multi-partition support. Called by /haipipe-task when task-type=data. Cross-references /haipipe-data for Fn…
Langue du texte source : anglais
Raw extraction task-folder specialist: scaffolds {NN}_<name>/ task-folders in the raw-extraction task-group (default R-series). Two patterns: extract-wide-process-local (Databricks -> parquet -> local Python; non-PHI) and server-resident (all-Spark pipeline…
Langue du texte source : anglais
Layer 1 (Algorithm) specialist of haipipe-nn: defines the algorithm contract (model class, forward pass, loss, metric). Covers mlpredictor, tsforecast, tefm, tediffusion, bandit. Called by /haipipe-nn; direct invocation works layer-scoped.
Langue du texte source : anglais
Layer 3 (Instance) specialist of haipipe-nn: materializes a trained ModelInstance by driving its Tuners (registry create -> fit -> save_model). Called by /haipipe-nn; direct invocation works layer-scoped.
Langue du texte source : anglais
Layer 4 (ModelSet / Pipeline) specialist of haipipe-nn: composes multiple ModelInstances into a registry-backed pipeline. Called by /haipipe-nn; direct invocation works layer-scoped.
Langue du texte source : anglais
Run any Stage 5 NN pipeline work: parses intent (layer + function) and dispatches to the right specialist (algo/tuner/instance/modelset). Use for algorithms (mlpredictor/tsforecast/tefm/tediffusion/bandit), tuner sweeps, ModelInstance materialization,…
Langue du texte source : anglais
Layer 2 (Tuner) specialist of haipipe-nn: defines the hyperparameter search space and the tuner that drives sweeps. Called by /haipipe-nn; direct invocation works layer-scoped.
Langue du texte source : anglais
algo-dev task-folder specialist: scaffolds {NN}_<name>/ task-folders in the algo-dev task-group (default X_algo) that smoke-test a newly developed algorithm class end-to-end on a TINY config. NOT full training -- see /haipipe-task-for-fit. Called by…
Langue du texte source : anglais
Databricks Model Serving deploy specialist for haipipe-end: wraps an Endpoint_Set into MLflow pyfunc + Unity Catalog model, deploys to Databricks Model Serving, runs live smoke tests, monitors, tears down. Reads (never modifies) Endpoint_Sets from…
Langue du texte source : anglais
Local self-hosted deploy specialist for haipipe-end: wraps an Endpoint_Set into a local HTTP server -- Flask (default), FastAPI, or local Docker container -- for dev, integration testing, demos, and DIY deployments. Reads (never modifies) Endpoint_Sets from…
Langue du texte source : anglais
MLflow deploy specialist for haipipe-end. STATUS: DEFERRED -- no platform-mlflow-inference repo backs it yet; placeholder. Would register an Endpoint_Set into an MLflow Model Registry and serve via `mlflow models serve`. Reads (never modifies) Endpoint_Sets…
Langue du texte source : anglais
AWS SageMaker deploy specialist for haipipe-end: wraps an Endpoint_Set into SageMaker model.tar.gz, deploys to a SageMaker endpoint, runs live smoke tests, monitors logs, tears down. Reads (never modifies) Endpoint_Sets from haipipe-end-endpointset. Read the…
Langue du texte source : anglais
Databricks develop specialist for haipipe-end. STATUS: DEFERRED -- backing repo platforms/platform-databrick-training/ exists but this skill isn't wired to it yet. Would run Stage 5 training as a Databricks Job with model logged to Unity Catalog and exported…
Langue du texte source : anglais
Local develop specialist for haipipe-end: runs Stage 5 training on the local machine and produces an Endpoint_Set under 6-EndpointStore/, for dev iteration, smoke tests, and DIY builds. Mostly delegates to /haipipe-nn modelset for training; exists for…
Langue du texte source : anglais
AWS SageMaker develop specialist for haipipe-end: runs Stage 5 training as a managed SageMaker Pipeline (Preprocess -> Train -> Reorganize -> RegisterModel) and produces a deployable Endpoint_Set / registered model package. Writes Endpoint_Sets that…
Langue du texte source : anglais
Endpoint_Set artifact-as-whole specialist: target-agnostic operations on the deployable artifact -- package (Stage 5 -> 6), local inference() smoke test, structural review, dashboard. Per-Fn-type design/review lives in…
Langue du texte source : anglais
Input2SrcFn specialist -- designs/reviews the wire-payload->record function in an Endpoint_Set (deserializes a JSON request into a ProcessedDF row). Platform-specific: one impl per deploy platform (SageMaker flat JSON vs Databricks dataframe_records);…
Langue du texte source : anglais
MetaFn specialist -- designs/reviews the model-metadata-lookup function in an Endpoint_Set. One of 5 inference Fn-types. Called by /haipipe-end when intent references MetaFn, model metadata, model card, or `meta`.
Langue du texte source : anglais
TrigFn specialist -- designs/reviews the trigger-detection function in an Endpoint_Set. One of 5 inference Fn-types. Called by /haipipe-end when intent references TrigFn, trigger detection, or `trig`.
Langue du texte source : anglais
endpoint task-folder specialist: scaffolds AND executes {NN}_<name>/ task-folders that package a trained ModelInstance_Set into a deployable Endpoint_Set (Stage 6) via c_endpoint_nb.py. Called by /haipipe-task when task-type=endpoint. Cross-references…
Langue du texte source : anglais