Results for “research”

Provides guidance for mechanistic interpretability research using TransformerLens to inspect and manipulate transformer internals via HookPoints and activation caching. Use when reverse-engineering model algorithms, studying attention patterns, or performing activation patching experiments.

Provides guidance for mechanistic interpretability research using TransformerLens to inspect and manipulate transformer internals via HookPoints and activation caching. Use when reverse-engineering model algorithms, studying attention patterns, or performing activation patching experiments.

Use when incorporating information from research, references, or external sources. This skill provides verification procedures to ensure information used in deliverables is accurate, current, and from authoritative sources, preventing hallucination and misinformation.

Search X/Twitter profiles, tweets, trends, and approved engagement actions through the AIsa relay. Use when the user asks for Twitter/X research, posting, likes, follows, or related workflows without sharing passwords.

Generate unique personalized first lines for hundreds of prospects using company news, LinkedIn activity, and mutual connections. Saves 10+ hours of manual research per campaign. Use when you need personalized outreach at volume.

Capture, organize, and retrieve notes efficiently using structured formats, tagging, and file management for meetings, ideas, research, and daily logs. Use when the user requests note taking or provides relevant inputs for this workflow.

Use this skill when the user requests to generate, create, or write professional research reports including but not limited to market analysis, consumer insights, brand analysis, financial analysis, industry research, competitive intelligence, investment due diligence, or any consulting-grade analytical report. This skill operates in two phases — (1) generating a structured analysis framework with chapter skeleton, data query requirements, and analysis logic, and (2) after data collection by other skills, producing the final consulting-grade report with structured narratives, embedded charts, and strategic insights.

Use this skill when a user needs KOL or influencer research, creator email lookup, similar-creator discovery, outreach-list building, influencer prospecting, or a contact table from TikTok, Instagram, or YouTube profile URLs. It uses AIsa's WaveInflu APIs to find verified creator emails, match similar YouTube or TikTok creators, enrich each recommended profile with contact emails, and return an outreach-ready Markdown table without inventing missing data. Use when: the user needs YouTube search, trend discovery, channel research, or SERP analysis.

Requires python3, and AISA_API_KEY. Uses the supplied AISA_API_KEY to send requests to https://api.aisa.one. Native-first ClawHub plugin for `multi-search`. Ships the packaged AIsa skill with an `openclaw.plugin.json` manifest and a Claude-compatible bundle fallback. Parallel multi-source search combining Web, Scholar, Smart, and Tavily results with confidence scoring and AI synthesis. Best for comprehensive research requiring cross-source validation. Use when: the user needs web search, research, source discovery, or content extraction.

Create sales-ready competitive battlecards comparing your product against a specific competitor, including positioning, feature comparison, objection handling, and win/loss patterns.

Build an Opportunity Solution Tree (OST) to structure product discovery — map a desired outcome to opportunities, solutions, and experiments.

Audit, plan, and implement SEO improvements across technical SEO, on-page optimization, structured data, Core Web Vitals, and content strategy.

Defines a gated Research-Plan-TDD-Review-Commit pipeline for orchestrating feature development, defect fixes, and refactoring tasks through composable phases and human approval gates.

Research any topic across Reddit, X, and the web from the last 30 days, then synthesize findings into actionable insights and copy-paste-ready prompts.

Extracts structured data from Google Search results pages, including organic listings, paid ads, related searches, People Also Ask, and AI Overview.

Extracts structured image metadata from Google Images via the BrowserAct API, including thumbnails, titles, source logos, and click-through URLs.

Extracts structured product data from Amazon search results, including prices, ratings, sales estimates, and shipping info, using the BrowserAct API.

Extracts structured channel data from YouTube search results via the BrowserAct API, including subscriber counts, verification status, and channel descriptions.

Quantize large language models to 4-bit using activation-aware weight quantization, achieving ~3x speedup with minimal accuracy loss for deployment on limited GPU memory.

Transcribe and translate speech across 99 languages using OpenAI's Whisper model, with support for multiple model sizes, batch processing, and subtitle generation.

Post-train LLMs with reinforcement learning using the slime framework, which integrates Megatron-LM for training and SGLang for rollout generation.

Build complex AI systems with declarative programming, optimize prompts automatically, and create modular RAG systems and agents using Stanford NLP's DSPy framework.

Process large-scale ML datasets with distributed streaming execution across CPU/GPU, supporting Parquet, CSV, JSON, images, and integration with PyTorch, TensorFlow, and Ray Train.

Provides expert guidance for fine-tuning LLMs with LLaMA-Factory, covering WebUI no-code, 100+ models, 2/3/4/5/6/8-bit QLoRA, and multimodal support.

Evaluates LLMs across 60+ academic benchmarks (MMLU, HumanEval, GSM8K, TruthfulQA, HellaSwag) using standardized prompts and metrics. Supports HuggingFace, vLLM, and API backends.

Trains large language models (2B-462B parameters) using NVIDIA Megatron-Core with advanced parallelism strategies for maximum GPU efficiency.

Analyze card sorting results to inform information architecture and navigation structure. Use after conducting open or closed card sort studies.

Researches competitor paid-ad presence, messaging, creative, formats, landing pages, and strategic gaps using official transparency libraries and public sources.

Guides authorized hardware and embedded interface security research including UART/JTAG discovery, debug pad triage, secure boot assessment, and offline firmware extraction support.

Extract and analyze Douyin video data into evidence-graded learning notes, summaries, and research briefs using subtitles, ASR, comments, and AI-assisted reading.

Backtests quantitative trading strategies across 9 engines and 25 data sources, analyzes trade journals, and runs multi-agent research teams.

Searches UK Biobank's 12,000+ data fields and publications by natural language query, returning ranked field IDs and descriptions for research questions.

Audits and improves search visibility through technical SEO, on-page optimization, structured data, Core Web Vitals, and content strategy.

Search X/Twitter profiles, tweets, trends, and approved engagement actions through the AIsa relay. Use when the user asks for Twitter/X research, posting, likes, follows, or related workflows without sharing passwords.

Plan new YouTube videos with optimized titles, thumbnails, and hooks. Use when planning a video, writing or reviewing opening hooks, generating title and thumbnail concepts, or creating a production-ready video plan backed by research.

Capture structured feedback on the Han skills and agents used in the current session and optionally post it as a GitHub issue to testdouble/han. Use at the end of any session where one or more han-* skills or agents ran, to rate a run, log what worked and what didn't, or submit observations for maintainers. Does not review code, investigate bugs, or research options; use code-review, investigate, or research for those. Does not provide feedback on skills or agents from non-Han plugins.