Researches a topic via web search and writes a cited Markdown memo in Japanese.
npx skills add https://github.com/mjun0812/skills --skill deep-research-mjun0812
Please help me install this Agent Skill. Skill: deep-research Source: https://github.com/mjun0812/skills/tree/main/skills/research/deep-research Command: npx skills add https://github.com/mjun0812/skills --skill deep-research-mjun0812
[n] citations to key claims, prioritizes primary sources, and follows bundled Japanese technical-writing and anti-AI-slop style references. - Optional saving with assets: Presents the memo for review, then saves it as YYYY-MM-DD_<slug>.md with frontmatter, downloads referenced figures into images/, and validates citation links; --no-save returns the memo inline for use from other skills or non-interactive runs. - Use Case: Ask "deep research して: RAG の評価手法の最新動向" and receive a conclusion-first Japanese memo with numbered references, unresolved questions, and an optional dated file saved to the current directory. ## Quick Start Ask the agent to deep research a topic and write a cited memo, for example: 「LLMエージェントの評価手法について deep research してメモにして」.High-intent search queries and answers about installing and using this skill.
Invoke the skill with a research theme, and it decomposes the topic into 3-8 sub-questions, searches the web in English and Japanese, reads primary sources, and writes a Markdown memo where key claims carry [n] citations mapped to a reference list.
The memo is a single Markdown document in Japanese with a conclusion-first structure, per-topic sections, an unresolved-questions section, and a numbered reference list. When saved, it becomes YYYY-MM-DD_<slug>.md with title, created, updated, and model frontmatter.
Yes, pass the --no-save argument. The skill then skips the save confirmation and file creation, returning the memo body directly as the response, which suits calls from other skills, agents, or claude -p style executions.
Yes, the skill explicitly avoids depending on vendor-specific tools. It instructs the agent to use whatever web search and parallel execution mechanisms are available in the current environment.
It only performs investigation and memo writing; it does not summarize, organize, or integrate the memo into other files. For narrow topics with few sources, it reports the scarcity rather than padding the source count.
smartfrogAsk 1-5 essential clarifying questions before starting ambiguous work.
mike2oxConstruct memos from conversation transcripts with web research insights.
FPXszkPolish Japanese social media posts using anti-AI writing rules.
kortix-aiConducts multi-source web research and produces cited, evidence-based reports.