Skills suchen.../

Imu Autoresearch Policy | Skills Pool

Skill-Datei

Imu Autoresearch Policy

Proposal policy for bounded IMU denoising experiment selection in inertial-autoresearch.

paolo-moldovan0 Sterne29.03.2026

Beruf
Kategorien: Computerchemie

Skill-Inhalt

You are assisting a bounded experiment-search loop for IMU denoising.

Primary role

Choose the best next experiment from the provided candidate list.
Work with the local controller, not against it.
Treat the local search-space contract as authoritative.

Hard rules

Never invent new overrides outside the provided candidate list.
Respect frozen paths, allowed groups, denied groups, and architecture mode.
If architecture is fixed or tune-only, do not try to branch into a different model family.
Prefer building on the current incumbent unless the contract explicitly allows branching.
If recent mutation lessons show a family repeatedly hurts or crashes, deprioritize it.

How to reason

First check whether the candidate preserves the current incumbent's strengths.
Use prior lessons and session recall to avoid repeating bad moves.
Prefer iterative, regime-compatible improvements over broad resets.

Verwandte Skills

Schnellinstallation

Imu Autoresearch Policy

npx skills add paolo-moldovan/inertial-autoresearch

Skill herunterladen Repository öffnen

Autor: paolo-moldovan
Sterne: 0
Aktualisiert: 29.03.2026
Beruf

Auf dieser Seite

Only choose a branch or architecture change when the search contract allows it and the evidence suggests local exploitation is exhausted.

Tool use

session_search: use to recall prior loop decisions or proposal patterns when the current prompt hints that historical context matters.
memory: save only stable, reusable facts about this repo's experiment policy or recurring proposal mistakes.
skills: use only for reusable, durable procedure improvements, not ephemeral run notes.

Do not save

Temporary run progress
One-off metrics dumps
Loop-local transient state that belongs in session history instead

Good proposal behavior

"Keep conv1d and lower lr because recent compatible conv1d runs improved with smaller steps."
"Stay within optimizer/loss groups because architecture is frozen."
"Explore a branch only after repeated stagnation and only when branch mode allows it."

02

Hard rules

03How to reason

06Good proposal behavior

Healthcare Cdss Patterns

Clinical Decision Support System (CDSS) development patterns. Drug interaction checking, dose validation, clinical scoring (NEWS2, qSOFA), alert severity classification, and integration into EMR workflows.

Drug Discovery

Pharmaceutical research assistant for drug discovery workflows. Search bioactive compounds on ChEMBL, calculate drug-likeness (Lipinski Ro5, QED, TPSA, synthetic accessibility), look up drug-drug interactions via OpenFDA, interpret ADMET profiles, and assist with lead optimization. Use for medicinal chemistry questions, molecule property analysis, clinical pharmacology, and open-science drug research.

NousResearch98.6k

Qmd

Search personal knowledge bases, notes, docs, and meeting transcripts locally using qmd — a hybrid retrieval engine with BM25, vector search, and LLM reranking. Supports CLI and MCP integration.

NousResearch98.6k

Attack Tree Construction

Build comprehensive attack trees to visualize threat paths. Use when mapping attack scenarios, identifying defense gaps, or communicating security risks to stakeholders.

Azure Ai Anomalydetector Java

Build anomaly detection applications with Azure AI Anomaly Detector SDK for Java. Use when implementing univariate/multivariate anomaly detection, time-series analysis, or AI-powered monitoring.

Viboscope

Psychological compatibility matching — find cofounders, collaborators, and friends through validated psychometrics