{"id":52615,"date":"2026-04-30T17:01:49","date_gmt":"2026-04-30T09:01:49","guid":{"rendered":"https:\/\/www.1ai.net\/?p=52615"},"modified":"2026-04-30T17:06:57","modified_gmt":"2026-04-30T09:06:57","slug":"hermes%e9%ab%98%e7%ba%a7%e7%94%a8%e6%b3%95%ef%bc%9a%e7%94%a8%e5%a4%9a-profile-%e5%8d%8f%e4%bd%9c-wiki-%e5%85%b1%e4%ba%ab%e8%ae%b0%e5%bf%86%ef%bc%8c%e6%90%ad%e5%bb%ba%e4%bd%a0%e7%9a%84-opc-agent","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/52615.html","title":{"rendered":"Hermes Advanced Use: Working with MultiProfile + Wiki Sharing Memory to Build Your OPC Agent Team (Previously)"},"content":{"rendered":"<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-52616\" title=\"bc4289c6j00teaucy006yd000iw007kp\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2026\/04\/bc4289c6j00teaucy006yd000iw007kp.jpg\" alt=\"bc4289c6j00teaucy006yd000iw007kp\" width=\"680\" height=\"272\" \/><\/p>\n<p>recent <a href=\"https:\/\/www.1ai.net\/en\/tag\/opc\" title=\"[SEE ARTICLES WITH [OPC] LABELS]\" target=\"_blank\" >OPC<\/a> The concept is hot\u3002<\/p>\n<p>One-Person Company, one company\u3002<\/p>\n<p>Its core is not one single-handedness, but one single person who uses multiple Agents to perform one or more complex tasks, like managing a small team\u3002<\/p>\n<p>So the question is:<\/p>\n<blockquote>\n<ul>\n<li>How to build a reliable multi-Profile collaborative architecture<\/li>\n<\/ul>\n<\/blockquote>\n<p>I'll show you the article today <a href=\"https:\/\/www.1ai.net\/en\/tag\/hermes\" title=\"_Other Organiser\" target=\"_blank\" >Hermes<\/a> Core logic of multiProfile collaboration\u3002<\/p>\n<p>This is not simply a \"a few more Agent\" but a stable, clear, long-term Agent working system\u3002<\/p>\n<p>EVERYONE WHO WANTS TO DO OPC SHOULD THINK ABOUT IT\u3002<\/p>\n<p><strong>I. WHY USE MORE PROFILE COORDINATION<\/strong><\/p>\n<p>An Agent can certainly do a lot of things\u3002<\/p>\n<p>It can search for information, write articles, write codes, make drawings, and make a flash drive\u3002<\/p>\n<p>But if all the tasks were given to an Agent, three problems would arise in the long run\u3002<\/p>\n<p>1. The problem of hallucinations<\/p>\n<p>The biggest question for Agent is:<\/p>\n<blockquote>\n<ul>\n<li>Write yourself, judge yourself, think you're okay\u3002<\/li>\n<\/ul>\n<\/blockquote>\n<p>So we need to build a system of team operations, and we need to get different people to take on different tasks and look at the same issue from different angles\u3002<\/p>\n<p>for example:<\/p>\n<ul>\n<li>Researcher is responsible for the facts<\/li>\n<li>Writer is responsible for expression<\/li>\n<li>Builder is responsible for this<\/li>\n<li>Coordinator is in charge of integration\u3002<\/li>\n<\/ul>\n<p>This makes it easier for the system to detect loopholes and to get into single Agent self-identification\u3002<\/p>\n<p>2. Avoiding contamination of memory<\/p>\n<p>An Agent simultaneously does all the projects, all the links, and memory is easily mixed\u3002<\/p>\n<p>For example, it also remembers:<\/p>\n<ul>\n<li>The article will tell stories<\/li>\n<li>The code needs to be operational in priority<\/li>\n<li>(B) MVPS FOR PRODUCTS<\/li>\n<li>The research should indicate the source\u3002<\/li>\n<\/ul>\n<p>These experiences are not wrong in themselves, but they are different mandates, different links, different projects\u3002<\/p>\n<p>if we all blend into one memory, it'll be polluting and contradictory\u3002<\/p>\n<p>The final results are:<\/p>\n<ul>\n<li>The code is written with content creation habits<\/li>\n<li>The article is written with the idea of engineering<\/li>\n<li>The study began earlier\u3002<\/li>\n<\/ul>\n<p>That's memory pollution\u3002<\/p>\n<p>3. Avoiding confusion about roles<\/p>\n<p>One, Agent, is also responsible for research, planning, writing, implementation, review, replaying, and can easily be confused\u3002<\/p>\n<p>Typical manifestations are:<\/p>\n<ul>\n<li>When the study began, it began to draw conclusions<\/li>\n<li>When writing, it rechecks the data<\/li>\n<li>At the time of the review, it began to defend its own output<\/li>\n<li>When it comes to project management, it is obsessed with detailed execution\u3002<\/li>\n<\/ul>\n<p>So, the essence of multiProfile is not \"more Agent,\" but a clearer role boundary\u3002<\/p>\n<p><strong>II. UNDERSTANDING THE FOUR CONTEXTS: PROFILE, SUBAGENT, PROJECT, Wiki<\/strong><\/p>\n<p>Before we build a team, we have four different concepts:<\/p>\n<ul>\n<li>Profile<\/li>\n<li>Subagent<\/li>\n<li>Project<\/li>\n<li>Wiki<\/li>\n<\/ul>\n<p>If the four concepts were mixed, the system behind them would be messy\u3002<\/p>\n<p>1. Profile: long-term staff<\/p>\n<p>Profile can be understood as a permanent employee in the system\u3002<\/p>\n<p>Each Profile is an independent Agent with its own identity, memory, skills and operational configuration\u3002<\/p>\n<p>for example:<\/p>\n<ul>\n<li>Coordinator is the coordinator<\/li>\n<li>Researcher is a researcher<\/li>\n<li>Writer is a writer<\/li>\n<li>Builder is the builder\u3002<\/li>\n<\/ul>\n<p>They are not different chat windows for the same Agent, but long-term members of different roles, different duties, different memory boundaries\u3002<\/p>\n<p>So, there's a lot of Profile actually building up an Agent team\u3002<\/p>\n<p>Subagent: Temporary workers<\/p>\n<p>Subagent is temporary Agent\u3002<\/p>\n<p>It's more like an ad hoc assistant who can handle local issues in complex missions\u3002<\/p>\n<p>For example, if you want to write an in-depth article, you can send a couple of sub-agents:<\/p>\n<ul>\n<li>A CHECK-UP OF TRADITION RAG<\/li>\n<li>A survey of Agenic RAG<\/li>\n<li>LLM Wiki<\/li>\n<li>A check for a logical loophole\u3002<\/li>\n<\/ul>\n<p>Subagent ends with no permanent personality or memory\u3002<\/p>\n<p>So it can simply be understood that:<\/p>\n<blockquote>\n<ul>\n<li>Profile is a permanent employee. Subagent is temporarily outsourced\u3002<\/li>\n<\/ul>\n<\/blockquote>\n<p>3. Project: Project space<\/p>\n<p>Project is the project space for a long-term mission\u3002<\/p>\n<p>If you advance several long-term tasks at the same time, for example:<\/p>\n<ul>\n<li>Twitter Growth<\/li>\n<li>Vibe Coding<\/li>\n<li>Content growth systems<\/li>\n<li>Product Demo\u3002<\/li>\n<\/ul>\n<p>Each long-term mission should have its own project space\u3002<\/p>\n<p>Notice:<\/p>\n<blockquote>\n<ul>\n<li>When running multiple long-term missions, the key is not to build multi-Profile, but to divide project space first\u3002<\/li>\n<\/ul>\n<\/blockquote>\n<p>Do not copy a package for each item\u3002<\/p>\n<p>The correct approach is:<\/p>\n<blockquote>\n<ul>\n<li>The same set of Project Teams that serve multiple projects\u3002<\/li>\n<\/ul>\n<\/blockquote>\n<p>So as to avoid the explosion of the Profile number and long-term memory pollution\u3002<\/p>\n<p>Wiki: Shared memory layer<\/p>\n<p>How does a complex task move forward<\/p>\n<p>The answer is:<\/p>\n<blockquote>\n<ul>\n<li>Share Wiki\u3002<\/li>\n<\/ul>\n<\/blockquote>\n<p>Wiki is like a corporate shared document\u3002<\/p>\n<p>Different employees have different brains, but they can synchronize through shared documents:<\/p>\n<ul>\n<li>Project progress<\/li>\n<li>Mission status<\/li>\n<li>Decision-making records<\/li>\n<li>Knowledge sank\u3002<\/li>\n<\/ul>\n<p>In Hermes MultiProfile, Wiki is not an ordinary notebook, but an organized, maintenanceable, long-term, shared memory system\u3002<\/p>\n<p>It records:<\/p>\n<ul>\n<li>Project status<\/li>\n<li>Progress of the mandate<\/li>\n<li>Decision-making records<\/li>\n<li>Research materials<\/li>\n<li>Final outputs<\/li>\n<li>General methodology\u3002<\/li>\n<\/ul>\n<p>So, a stable Agent team is actually made up of multiple Profiles and shared Wikis\u3002<\/p>\n<p><strong>III. The Four Role Model: Organizing like a real company<\/strong><\/p>\n<p>What about the division of labour<\/p>\n<p>I recommend a four-player model:<\/p>\n<ul>\n<li>Cordinator<\/li>\n<li>Researcher<\/li>\n<li>Writer<\/li>\n<li>Builder<\/li>\n<\/ul>\n<p>It's understandable:<\/p>\n<ul>\n<li>Project Manager<\/li>\n<li>researcher<\/li>\n<li>Writer<\/li>\n<li>engineer<\/li>\n<\/ul>\n<p>Role 1: Coordinator<\/p>\n<p>Cordinator is the project manager\u3002<\/p>\n<p>Its core functions are:<\/p>\n<ol>\n<li>Definition of objectives: Clarify what is ultimately to be achieved\u3002<\/li>\n<li>Disaggregation of tasks: Distinction of targets into implementable tasks\u3002<\/li>\n<li>Route tasks: judge to whom each task should be entrusted\u3002<\/li>\n<li>Summarized results: integration of outputs from different actors into final results\u3002<\/li>\n<li>Checking borders: Avoiding memory pollution and document conflict\u3002<\/li>\n<\/ol>\n<p>Cordinator should not be obsessed with personal research, writing, writing\u3002<\/p>\n<p>Its most important functions are:<\/p>\n<blockquote>\n<ul>\n<li>Let the whole system run in order\u3002<\/li>\n<\/ul>\n<\/blockquote>\n<p>Role 2: Researcher<\/p>\n<p>Researcher is a researcher\u3002<\/p>\n<p>Its core functions are:<\/p>\n<ol>\n<li>Collection of evidence: information from multiple sources\u3002<\/li>\n<li>Source of comparison: Cross-check to ensure reliable information\u3002<\/li>\n<li>Marking Uncertainty: Specify which information has not been validated\u3002<\/li>\n<li>Refinement of facts: Distinguishing facts, views and assumptions\u3002<\/li>\n<\/ol>\n<p>A good Researcher can significantly reduce the illusion of the system\u3002<\/p>\n<p>It should not write the final draft or make the final decision\u3002<\/p>\n<p>What it needs to do is to provide reliable raw materials\u3002<\/p>\n<p>Role 3: Writer<\/p>\n<p>Writer is responsible for translating raw materials into clear content\u3002<\/p>\n<p>Its core functions are:<\/p>\n<ol>\n<li>Sets the structure of the article\u3002<\/li>\n<li>Optimizing expression\u3002<\/li>\n<li>Refining main lines and perspectives\u3002<\/li>\n<li>It's a good way to make content fit for the target audience\u3002<\/li>\n<li>Clarify the concept of complexity\u3002<\/li>\n<\/ol>\n<p>When Writer is no longer responsible for planning and checking information, its output quality will increase significantly\u3002<\/p>\n<p>Because it can focus on one thing:<\/p>\n<blockquote>\n<ul>\n<li>Make it clear\u3002<\/li>\n<\/ul>\n<\/blockquote>\n<p>Role 4: Builder Builder<\/p>\n<p>Builder is more like an engineer\u3002<\/p>\n<p>Its core functions are:<\/p>\n<ol>\n<li>Achieved: the plan is translated into a running code, page or system\u3002<\/li>\n<li>Debugging: Positioning and repairing problems\u3002<\/li>\n<li>Test: Ensure that the output is stable and reliable\u3002<\/li>\n<li>Delivery: Generate final available results\u3002<\/li>\n<\/ol>\n<p>When Builder is no longer responsible for telling stories, doing research, or setting directions, its achievement will also improve\u3002<\/p>\n<p>Because it can focus on landing\u3002<\/p>\n<p><strong>IV. Full workflow<\/strong><\/p>\n<p>A typical set of multipleProfile streams:<\/p>\n<ol>\n<li>Coordinator decomposition and mission planning<\/li>\n<li>Researcher collects sources, validates claims<\/li>\n<li>Writer translates research findings into clear content<\/li>\n<li>Builder is responsible for final realization or delivery<\/li>\n<li>Cordinator final check, summarize, archive\u3002<\/li>\n<\/ol>\n<p>This structure is strong because it reflects real work processes\u3002<\/p>\n<p>In reality, a stabilization team does not allow the same person to be project manager, researcher, writer, engineer and reviewer at the same time\u3002<\/p>\n<p>The same is true of the Agent team\u3002<\/p>\n<p><strong>v. Starting construction of each file<\/strong><\/p>\n<p>When the role division is determined, you can start building each file\u3002<\/p>\n<p>EachProfile usually contains:<\/p>\n<p>soul.mdUSER.mdmemory.mdconfig.yamlskills\/.env<\/p>\n<p>Seeing so many things, a lot of people will be big\u3002<\/p>\n<p>Next, I would like to take the example of Coordinator, who is responsible for each document\u3002<\/p>\n<p>1. soul.md: who is the coordinator<\/p>\n<p>soul.md is the core identity document for Profile\u3002<\/p>\n<p>PROFILE:<\/p>\n<ul>\n<li>Who is it<\/li>\n<li>Responsible for what<\/li>\n<li>What borders are there<\/li>\n<li>Shouldn't do anything\u3002<\/li>\n<\/ul>\n<p>For example, Coordinator's soul.md needs clarification:<\/p>\n<ul>\n<li>It is the coordinator<\/li>\n<li>It is responsible for the separation of tasks, the planning of projects and the aggregation of results<\/li>\n<li>it maintains dashboard and angent-log<\/li>\n<li>It does not carry out research missions directly<\/li>\n<li>It does not write the final content directly<\/li>\n<li>It doesn't happen directly to the code\u3002<\/li>\n<\/ul>\n<p>Notice:<\/p>\n<blockquote>\n<ul>\n<li>if you have multiple projects at the same time, do not include specific items in soul.md\u3002<\/li>\n<\/ul>\n<\/blockquote>\n<p>Otherwise, it is highly susceptible to role pollution\u3002<\/p>\n<p>soul.md writes:<\/p>\n<blockquote>\n<ul>\n<li>This Agent is who\u3002<\/li>\n<\/ul>\n<\/blockquote>\n<p>No:<\/p>\n<blockquote>\n<ul>\n<li>What is this project doing now\u3002<\/li>\n<\/ul>\n<\/blockquote>\n<p>2. USER.md: Who is the user as understood by the Coordinator<\/p>\n<p>USER.md records this Profile understanding of users\u3002<\/p>\n<p>for example:<\/p>\n<ul>\n<li>User preference for communication in Chinese<\/li>\n<li>Users like a clearly structured Markdown<\/li>\n<li>Users do not like generic concepts<\/li>\n<li>The user wants the output to be suitable for copying to Obsidian\u3002<\/li>\n<\/ul>\n<p>Notice:<\/p>\n<p>The USER.md here is the \"user image understood by the coordinator role\" and is not the only user image of the system\u3002<\/p>\n<p>If all Profile shared user images are needed, they should be placed in Wiki:<\/p>\n<p>system\/user-profile.md<\/p>\n<p>3. memoory.md: common lessons learned by coordinators<\/p>\n<p>the memory.md records the generic experience that Profile has learned in the long term\u3002<\/p>\n<p>for example:<\/p>\n<ul>\n<li>Complex tasks should be dismantled<\/li>\n<li>Not to modify the same official document at the same time<\/li>\n<li>inbox<\/li>\n<li>final output re-enters outputs\u3002<\/li>\n<\/ul>\n<p>But there should be no project-specific experience here\u3002<\/p>\n<p>for example:<\/p>\n<ul>\n<li>Twitter wrote three tweets today<\/li>\n<li>Vibe Coding login page completed\u3002<\/li>\n<\/ul>\n<p>these elements do not belong to memory.md but should be placed in the project space\u3002<\/p>\n<p>i'm sorry.md recorded:<\/p>\n<blockquote>\n<ul>\n<li>Common experience, not project status\u3002<\/li>\n<\/ul>\n<\/blockquote>\n<p>4. skills: a pool of facilitators ' skills<\/p>\n<p>skills is a reusable task process\u3002<\/p>\n<p>Cordinator's skills can include:<\/p>\n<ul>\n<li>Dismantling of tasks<\/li>\n<li>Project priority judgement<\/li>\n<li>(a) The production of hand-over orders<\/li>\n<li>dashboard update<\/li>\n<li>i'm sorry<\/li>\n<li>i'm sorry\u3002<\/li>\n<\/ul>\n<p>these skills specifically serve the role of coordinator\u3002<\/p>\n<p>if some skill is a common skill, like:<\/p>\n<ul>\n<li>project-content-loader<\/li>\n<li>i'm sorry\u3002<\/li>\n<\/ul>\n<p>You can add multiple files list each\u3002<\/p>\n<p>But don't put all the skills into all the Profile\u3002<\/p>\n<p>Otherwise, the role boundaries will be disrupted\u3002<\/p>\n<p>5. config.yaml: how the coordinator works<\/p>\n<p>config.yaml is the running configuration for Profile\u3002<\/p>\n<p>It defines how this Profile works, for example:<\/p>\n<ul>\n<li>What models are used<\/li>\n<li>What is the default work directory<\/li>\n<li>Which documents are allowed to read and write<\/li>\n<li>whether to automatically load skills<\/li>\n<li>Which documents cannot be modified\u3002<\/li>\n<\/ul>\n<p>Notice:<\/p>\n<blockquote>\n<ul>\n<li>config.yaml is not the place to write about the mission\u3002<\/li>\n<\/ul>\n<\/blockquote>\n<p>It replied:<\/p>\n<blockquote>\n<ul>\n<li>How does this Profile run<\/li>\n<\/ul>\n<\/blockquote>\n<p>No:<\/p>\n<blockquote>\n<ul>\n<li>What's this project gonna do<\/li>\n<\/ul>\n<\/blockquote>\n<p>6. .env: coordinator ' s key<\/p>\n<p>.env only put the key\u3002<\/p>\n<p>for example:<\/p>\n<ul>\n<li>API Key<\/li>\n<li>Token<\/li>\n<li>SMTP PASSWORD<\/li>\n<li>External service vouchers\u3002<\/li>\n<\/ul>\n<p>.env:<\/p>\n<ul>\n<li>Project status<\/li>\n<li>User preferences<\/li>\n<li>Mission statement<\/li>\n<li>Articles\u3002<\/li>\n<\/ul>\n<p>And don't write in:<\/p>\n<ul>\n<li>i'm sorry<\/li>\n<li>USER.md<\/li>\n<li>i'm sorry<\/li>\n<li>AGENTS.md<\/li>\n<li>Wiki<\/li>\n<\/ul>\n<p>Because these files may enter the context of the model\u3002<\/p>\n<p>Previous Summary<\/p>\n<p>Here, the basic setup of the multiple Profile is over\u3002<\/p>\n<p>We did three things:<\/p>\n<ol>\n<li>To make it clear why an Agent can't have a full package<\/li>\n<li>Distinguishing the concepts of Profile, Subagent, Project, Wiki<\/li>\n<li>The role models of Coordinator, Researcher, Writer, Builder\u3002<\/li>\n<\/ol>\n<p>But it's just a team level\u3002<\/p>\n<p>It is not enough for a team to run on a permanent basis\u3002<\/p>\n<p>It also requires:<\/p>\n<ul>\n<li>Shared documents<\/li>\n<li>Project space<\/li>\n<li>Mission board<\/li>\n<li>Decision-making records<\/li>\n<li>Rewind mechanisms\u3002<\/li>\n<\/ul>\n<p>This is the point of the next issue:<\/p>\n<blockquote>\n<ul>\n<li><a href=\"https:\/\/www.1ai.net\/en\/52617.html\/\">How to build your OPC shared memory system with Wiki<\/a>.<\/li>\n<\/ul>\n<\/blockquote>","protected":false},"excerpt":{"rendered":"<p>Recently, the concept of the OPC has been hot. One-Person Company, one company. Its core is not one single-handedness, but one single person who uses multiple Agents to perform one or more complex tasks, like managing a small team. So the question is: How to build a reliable, multi-Profile collaborative architecture? Today's article, I'm going to show you the core logic of Hermes' multi-Profile collaboration. This is not simply a \"a few more Agent\" but a stable, clear, long-term Agent working system. Everyone who wants to do OPC should think about it. I. WHY USE<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[144],"tags":[8462,8480,8009],"collection":[],"class_list":{"0":"post-52615","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"hentry","6":"category-baike","7":"tag-hermes","9":"tag-opc"},"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/52615","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=52615"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/52615\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=52615"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=52615"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=52615"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=52615"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}