{"id":"2061601347271577630","url":"https://x.com/cyrilXBT/status/2061601347271577630","text":"","author":{"name":"CyrilXBT","username":"cyrilXBT","avatarUrl":"https://pbs.twimg.com/profile_images/2035229727414534145/aWap3Jbq_200x200.jpg"},"createdAt":"Tue Jun 02 00:10:38 +0000 2026","engagement":{"replies":22,"retweets":62,"likes":405,"views":285900},"article":{"title":"How to Connect Obsidian + Hermes Agent Into One System That Thinks, Remembers, and Runs Your Life ","previewText":"Most productivity systems have a memory problem.\nYour Obsidian vault stores everything you know but cannot act on it without you initiating something.\nYour AI agent can act on anything but forgets","coverImageUrl":"https://pbs.twimg.com/media/HJsuN8OWUAUMria.jpg","content":"Most productivity systems have a memory problem.\n\nYour Obsidian vault stores everything you know but cannot act on it without you initiating something.\n\nYour AI agent can act on anything but forgets everything the moment the session closes.\n\nConnect them and the problem disappears entirely.\n\nObsidian becomes the permanent knowledge layer. Every note, every decision, every captured idea, every project update lives there as plain text that never disappears.\n\nHermes Agent becomes the intelligence layer that reads that knowledge, reasons across it, executes workflows against it, and writes new knowledge back into it automatically.\n\nTogether they form a system that thinks with your accumulated knowledge, remembers everything across every session, and runs your most important recurring workflows while you sleep.\n\nThis guide is the complete build.\n\n## Why This Combination Works\n\nBefore the architecture understand why these two tools belong together.\n\nThe problem with Obsidian alone:\n\nObsidian is the best knowledge storage system available. Plain text files. Local storage. Permanent. Searchable. Linkable. But Obsidian is passive. It stores what you put in and returns it when you search for it. It does not read your notes and surface insights. It does not run workflows. It does not generate outputs from your accumulated knowledge without you manually asking for them.\n\nThe problem with Hermes Agent alone:\n\nHermes is the best autonomous agent framework available. Persistent memory. Skill execution. Scheduled automation. MCP connections to real tools. But Hermes memory lives in a SQLite database. It is not human-readable. You cannot open it, browse it, or edit it directly. You cannot build a rich knowledge network inside it the way you can in Obsidian.\n\nWhat the combination produces:\n\nObsidian handles human-readable permanent storage. You read it, write in it, link notes together, and build a knowledge network that compounds over years.\n\nHermes reads your Obsidian vault via the Filesystem MCP, reasons across everything in it using Claude, executes automated workflows that produce new knowledge, and writes the outputs back into your vault as new notes.\n\nThe vault is not just where you store things. It is the operating environment that Hermes runs against.\n\nEvery note you write becomes context Hermes uses.\n\nEvery output Hermes generates becomes a note you can read, edit, and link.\n\nThe two systems compound together rather than operating in parallel.\n\n## The Architecture\n\nThe system has four layers that work together.\n\nLayer 1: The Vault Layer — Obsidian\n\nPlain text Markdown files organized in a structure designed for both human navigation and automated processing. Every note you write, every idea you capture, every decision you make lives here.\n\nLayer 2: The Connection Layer — Filesystem MCP\n\nThe bridge between Obsidian and Hermes. The Filesystem MCP gives Hermes direct read and write access to your Obsidian vault. Hermes can read any note, search across all notes, create new notes, and update existing ones.\n\nLayer 3: The Intelligence Layer — Hermes + Claude\n\nHermes executes skill files that call Claude via the Anthropic API. Claude reads your vault, reasons across your accumulated knowledge, and produces outputs that go back into the vault as new notes.\n\nLayer 4: The Automation Layer — Hermes Scheduler\n\nScheduled workflows that fire automatically at configured times. Morning briefings generated from your vault before you wake up. Weekly reviews synthesized every Sunday evening. Project status updates produced every Monday morning. All of it running without you initiating anything.\n\n## Setting Up the Connection\n\nStep 1: Install Hermes Agent\n\ngit clone https://github.com/hermes-agent/hermes\ncd hermes\nnpm install\n\nStep 2: Configure Claude as the model\n\nOpen .env and configure:\n\nMODEL_PROVIDER=anthropic\nMODEL_NAME=claude-opus-4-8\nANTHROPIC_API_KEY=your-anthropic-api-key\n\nMEMORY_BACKEND=sqlite\nMEMORY_PATH=./data/memory.db\n\nENABLE_SCHEDULER=true\nSCHEDULER_TIMEZONE=America/New_York\n\nStep 3: Configure the Filesystem MCP to point at your Obsidian vault\n\nMCP_FILESYSTEM_PATH=/Users/yourname/Documents/YourVaultName\n\nThis single configuration is what connects everything. Hermes now has direct access to every note in your vault.\n\nStep 4: Install and configure the MCP server\n\nnpm install -g @modelcontextprotocol/server-filesystem\nnpm install -g @modelcontextprotocol/server-brave-search\n\nStep 5: Start Hermes and verify the connection\n\nnpm run start\n\nTest the connection:\n\nList the top-level folders in my Obsidian vault \nand describe the structure you see.\n\nHermes should accurately describe your vault's folder structure. The connection is live.\n\n## The Vault Structure for the Combined System\n\nThe vault structure needs to accommodate both human navigation and automated processing. Hermes needs to know where to read inputs and where to write outputs. You need to know where to find everything Hermes produces.\n\nVAULT/\n    00 - INBOX/\n        [raw captures that need processing]\n    \n    01 - NOTES/\n        permanent/\n            [atomic notes in your own words]\n        daily/\n            [YYYY-MM-DD].md\n        meetings/\n            [YYYY-MM-DD]-[meeting-name].md\n    \n    02 - PROJECTS/\n        [project-name]/\n            overview.md\n            tasks.md\n            notes/\n    \n    03 - RESOURCES/\n        [topic-based reference notes]\n    \n    04 - HERMES-OUTPUTS/\n        briefings/\n            [YYYY-MM-DD]-morning-briefing.md\n        analyses/\n            [YYYY-MM-DD]-[topic]-analysis.md\n        syntheses/\n            [YYYY-MM-DD]-[topic]-synthesis.md\n        reviews/\n            [YYYY-MM-DD]-weekly-review.md\n    \n    05 - ARCHIVE/\n        [completed projects and outdated material]\n    \n    06 - SYSTEM/\n        CLAUDE.md\n        skills/\n        templates/\n\nThe HERMES-OUTPUTS folder is the critical addition. Every automated output Hermes generates lands here. You always know where to find what Hermes produced. You can read it, edit it, link it to other notes, and build on it.\n\nThe SYSTEM folder contains your CLAUDE.md and optionally copies of your Hermes skill files for reference. Keeping skill documentation in the vault means your notes about how the system works live alongside the knowledge the system produces.\n\n## The CLAUDE.md That Powers the System\n\nThe CLAUDE.md lives in your Obsidian vault at 06 - SYSTEM/CLAUDE.md. Hermes reads it via the Filesystem MCP at the start of every skill execution.\n\n# Obsidian + Hermes System — CLAUDE.md\n\n## Vault Location\nThis file is at: 06 - SYSTEM/CLAUDE.md\nAll outputs save to: 04 - HERMES-OUTPUTS/\n\n## Who I Am\nName: [YOUR NAME]\nWhat I do: [YOUR PRIMARY WORK]\nPrimary focus: [YOUR MOST IMPORTANT AREA]\n\n## How My Vault Is Organized\n00 - INBOX: Raw unprocessed captures\n01 - NOTES: Processed permanent notes and daily notes\n02 - PROJECTS: Active project folders\n03 - RESOURCES: Reference material by topic\n04 - HERMES-OUTPUTS: All automated outputs\n05 - ARCHIVE: Completed and outdated material\n06 - SYSTEM: CLAUDE.md and system files\n\n## Active Projects\n[PROJECT 1]: [ONE SENTENCE STATUS]\nNext action: [SPECIFIC NEXT STEP]\n\n[PROJECT 2]: [ONE SENTENCE STATUS]\nNext action: [SPECIFIC NEXT STEP]\n\n## Current Priorities\n1. [MOST IMPORTANT THING THIS WEEK]\n2. [SECOND MOST IMPORTANT]\n3. [THIRD MOST IMPORTANT]\n\n## Topics I Think About Most\n[LIST 5-8 RECURRING TOPICS]\n\n## Writing Voice\n[HOW YOU WRITE AND COMMUNICATE]\n[Specific things you never say or do]\n\n## Output Instructions\nWhen generating any output:\n- Read the relevant section of my vault first\n- Reference specific notes by their file path\n- Save outputs to the correct subfolder \n  in 04 - HERMES-OUTPUTS\n- Date-stamp every file: YYYY-MM-DD-[type]-[topic].md\n- Create proper Obsidian frontmatter with type and date\n\n## Memory Instructions\nBefore any skill execution:\n- Read relevant memory from Hermes database\n- Read relevant notes from vault via Filesystem MCP\n- Combine both sources for full context\n\nAfter any skill execution:\n- Store summary in Hermes memory with appropriate tags\n- Create or update the relevant vault note\n\n## Weekly Update\nThis file is reviewed and updated every Monday.\nCurrent week priorities updated: [DATE]\n\n## The Seven Skills That Make the System Work\n\nSkill 1: The Vault-Aware Morning Brief\n\nThis skill reads your actual Obsidian vault to generate a morning brief that is specific to your projects, your notes, and your current priorities rather than generic news.\n\nCreate skills/vault-morning-brief.md:\n\n# vault-morning-brief\n\n## Purpose\nGenerate a morning brief using my actual Obsidian \nvault as the primary context, supplemented by \nreal-time web search for external developments.\n\n## Trigger\nScheduled daily at 6:00 AM.\nManual: \"Generate my morning brief\"\n\n## Process\n1. Read CLAUDE.md from my vault at \n   06 - SYSTEM/CLAUDE.md\n\n2. Read today's daily note if it exists at\n   01 - NOTES/daily/[TODAY'S DATE].md\n\n3. Read all project overview.md files in \n   02 - PROJECTS/ to understand current project status\n\n4. For each active project in CLAUDE.md:\n   Check if any tasks are overdue by reading \n   the project's tasks.md file\n\n5. Read the most recent weekly review from \n   04 - HERMES-OUTPUTS/reviews/ to understand \n   last week's priorities and outcomes\n\n6. Use Brave Search to find external developments \n   relevant to my current priorities from CLAUDE.md\n\n7. Synthesize everything into the morning brief:\n\n---\n# Morning Brief — [DATE]\n\n## MOST IMPORTANT TODAY\n[Single highest-leverage action based on \nvault context and current priorities]\n\n## PROJECT PULSE\nFor each active project:\n[PROJECT NAME]: [One sentence status] \nNext: [Specific next action from tasks.md]\n\n## OPEN LOOPS\n[Any items from yesterday's daily note \nor recent notes marked as unfinished]\n\n## EXTERNAL SIGNAL\n[1-3 external developments relevant to \nmy current priorities]\n\n## FROM MY VAULT\n[One connection between something in my \nnotes and something happening today]\n---\n\n8. Save to: 04 - HERMES-OUTPUTS/briefings/[DATE]-morning-brief.md\n\n9. Create today's daily note at \n   01 - NOTES/daily/[DATE].md if it does not exist,\n   with a link to the morning brief\n\n## Output Format\nThe brief should reference specific note paths \nin my vault. Not generic observations. \nSpecific connections to actual notes I have written.\n\nSkill 2: The Inbox Processor\n\nThis skill empties your INBOX folder every evening, filing notes into the correct location and creating links to related permanent notes.\n\nCreate skills/inbox-processor.md:\n\n# inbox-processor\n\n## Purpose\nProcess all files in 00 - INBOX, file them \nin the correct vault location, and create \nconnections to related existing notes.\n\n## Trigger\nScheduled daily at 8:00 PM.\nManual: \"Process my inbox\"\n\n## Process\n1. Read CLAUDE.md for vault structure and context.\n\n2. List all files currently in 00 - INBOX\n\n3. For each file in INBOX:\n\n   a. READ the file content fully\n   \n   b. CLASSIFY the note type:\n      - PERMANENT IDEA: deserves its own atomic note\n      - PROJECT NOTE: belongs in a specific project\n      - REFERENCE: belongs in 03 - RESOURCES\n      - DAILY CAPTURE: belongs in today's daily note\n      - TASK: belongs in a project's tasks.md\n   \n   c. FIND CONNECTIONS: Search the vault for \n      existing notes that relate to this capture.\n      Look in 01 - NOTES/permanent/ primarily.\n   \n   d. PROCESS the note:\n      \n      For PERMANENT IDEA:\n      Create a new note at \n      01 - NOTES/permanent/[concept-name].md\n      with proper frontmatter and links to \n      related notes found in step c.\n      \n      For PROJECT NOTE:\n      Append to the relevant project's notes \n      or create a new note in \n      02 - PROJECTS/[project-name]/notes/\n      \n      For REFERENCE:\n      Create or append to the relevant topic \n      file in 03 - RESOURCES/\n      \n      For DAILY CAPTURE:\n      Append to 01 - NOTES/daily/[TODAY].md\n      \n      For TASK:\n      Append to the relevant project's tasks.md\n   \n   e. ARCHIVE the original inbox file to \n      05 - ARCHIVE/inbox-processed/\n\n4. Generate a processing summary:\n   [N] notes processed\n   [N] permanent notes created\n   [N] project notes filed\n   [N] connections made\n\n5. Save summary to: \n   04 - HERMES-OUTPUTS/[DATE]-inbox-summary.md\n\n## Quality Standard\nEvery permanent note created should have \nat least one link to an existing note. \nA permanent note with no connections is \nnot ready for the permanent folder.\n\nSkill 3: The Project Health Monitor\n\nThis skill reads every active project folder in your vault and produces a health assessment every Monday morning.\n\nCreate skills/project-health.md:\n\n# project-health\n\n## Purpose\nRead all active project folders and produce \na comprehensive project health report.\n\n## Trigger\nScheduled every Monday at 7:00 AM.\nManual: \"Check project health\"\n\n## Process\n1. Read CLAUDE.md for list of active projects.\n\n2. For each active project in 02 - PROJECTS/:\n   \n   a. Read overview.md\n   b. Read tasks.md\n   c. Read all notes in the project's notes/ folder\n   d. Check date of last modification on any file \n      in the project folder\n\n3. For each project Claude assesses:\n\n   STATUS: On Track / At Risk / Stalled / Blocked\n   \n   EVIDENCE: What specifically in the project files \n   indicates this status. Reference actual file content.\n   \n   NEXT ACTION: Single most important thing this \n   project needs in the next 48 hours.\n   \n   STALL FLAG: If no file in the project folder \n   has been modified in 7 or more days flag as stalled.\n\n4. Generate the health report:\n\n---\n# Project Health Report — [DATE]\n\n## Summary\n[N] active projects: [N] on track, [N] at risk, [N] stalled\n\n## Project Status\n\n### [PROJECT NAME]\nStatus: [STATUS]\nEvidence: [SPECIFIC REFERENCE TO VAULT CONTENT]\nNext action: [SPECIFIC ACTION]\n[Repeat for each project]\n\n## Attention Required\n[Any project flagged as stalled or blocked \nwith specific recommendation]\n---\n\n5. Save to: 04 - HERMES-OUTPUTS/reviews/[DATE]-project-health.md\n\n6. For any stalled project: add a REVIEW NEEDED \n   note to the project's overview.md with today's date\n\n## Note\nReference specific file paths and note content \nin the assessment. Generic observations about \nproject status are not useful. The report should \nonly say things that can be traced to specific \nnotes in the vault.\n\nSkill 4: The Note Connection Finder\n\nOnce per week this skill reads all notes created or modified in the past seven days and finds connections to older notes that were not made at creation time.\n\nCreate skills/connection-finder.md:\n\n# connection-finder\n\n## Purpose\nFind connections between recent notes and \nolder permanent notes that were not explicitly \nlinked at creation time.\n\n## Trigger\nScheduled every Sunday at 5:00 PM.\nManual: \"Find note connections\"\n\n## Process\n1. List all files in 01 - NOTES/ modified \n   in the past 7 days.\n\n2. For each recently modified note:\n   \n   a. Read the full note content\n   \n   b. Search all permanent notes in \n      01 - NOTES/permanent/ for semantic \n      connections to this note's content\n   \n   c. Identify connections that are NOT \n      already present as wikilinks in the note\n   \n   d. For each new connection found:\n      - Name both notes\n      - Describe specifically how they connect\n      - Assess connection strength: \n        Strong / Moderate / Weak\n      - Suggest the link text to use\n\n3. Filter to only Strong and Moderate connections.\n\n4. Generate the connection report:\n\n---\n# Connection Report — [DATE]\n\n## New Connections Found\n[N] connections across [N] notes\n\n### [NOTE NAME]\nPath: [FILE PATH]\nConnects to: [[OTHER NOTE NAME]]\nReason: [SPECIFIC DESCRIPTION OF CONNECTION]\nSuggested link text: [HOW TO PHRASE THE LINK]\n[Repeat for each connection]\n\n## Recommended Actions\nAdd these links this week to strengthen \nthe knowledge network.\n---\n\n5. Save to: 04 - HERMES-OUTPUTS/analyses/[DATE]-connections.md\n\n## Quality Standard\nOnly surface non-obvious connections. If the \nlink is already present or the connection is \ntrivially obvious from the note titles skip it. \nThe value is in finding the connections a human \nreader would miss.\n\nSkill 5: The Weekly Vault Synthesis\n\nEvery Sunday evening this skill reads your entire week of activity in the vault and produces a synthesis that reveals insights visible only across seven days of accumulated notes.\n\nCreate skills/weekly-synthesis.md:\n\n# weekly-synthesis\n\n## Purpose\nSynthesize the full week's vault activity \ninto a comprehensive review that surfaces \npatterns invisible in any individual day.\n\n## Trigger\nScheduled every Sunday at 7:00 PM.\nManual: \"Generate weekly synthesis\"\n\n## Process\n1. Read CLAUDE.md for current priorities \n   and active projects.\n\n2. Read all daily notes from the past 7 days \n   in 01 - NOTES/daily/\n\n3. Read all permanent notes created or \n   modified this week in 01 - NOTES/permanent/\n\n4. Read all Hermes outputs from this week \n   in 04 - HERMES-OUTPUTS/\n\n5. Read the project health report from Monday \n   in 04 - HERMES-OUTPUTS/reviews/\n\n6. Claude synthesizes across all of it:\n\n---\n# Weekly Synthesis — [WEEK OF DATE]\n\n## THE WEEK IN ONE LINE\n[Single sentence capturing the most \nimportant thing this week revealed]\n\n## WHAT MOVED FORWARD\n[Specific progress on active projects \nwith references to vault evidence]\n\n## WHAT DID NOT MOVE\n[Honest assessment of stalled items \nand the most likely reason]\n\n## IDEAS THAT EMERGED\n[New permanent notes created this week \nthat deserve attention]\n\n## THE PATTERN\n[One theme that appeared repeatedly \nacross the week's notes]\n\n## CONNECTIONS MADE\n[Most significant connections found \nbetween notes this week]\n\n## NEXT WEEK PRIORITIES\n[Three specific priorities for next week \nranked by impact, each with a next action]\n\n## CLAUDE.MD UPDATE NEEDED\n[Any priority or project status changes \nthat should be reflected in CLAUDE.md]\n---\n\n7. Save to: 04 - HERMES-OUTPUTS/reviews/[DATE]-weekly-synthesis.md\n\n8. Update CLAUDE.md Current Priorities section \n   with next week's top 3.\n\n## Note\nThe synthesis should be the document you read \non Monday morning before anything else. \nIt should orient you to the week ahead using \nthe full context of the week that just ended.\n\nSkill 6: The Research-to-Notes Converter\n\nWhen you drop research material into INBOX, this skill converts it into properly structured permanent notes with connections to your existing knowledge.\n\nCreate skills/research-converter.md:\n\n# research-converter\n\n## Purpose\nConvert research material dropped in INBOX \ninto properly structured permanent notes \nconnected to existing vault knowledge.\n\n## Trigger\nManual: \"Convert research in inbox\" or \n\"Process research on [TOPIC]\"\n\n## Process\n1. Read CLAUDE.md for vault context and \n   current priorities.\n\n2. Identify research material in 00 - INBOX \n   by reading file contents and looking for \n   source URLs, article text, or research notes.\n\n3. For each research item:\n\n   a. READ the full content\n   \n   b. EXTRACT the key ideas:\n      - What is the core claim?\n      - What is the supporting evidence?\n      - What is genuinely new versus \n        what I likely already know?\n   \n   c. SEARCH my existing vault for \n      01 - NOTES/permanent/ notes on related topics\n   \n   d. CREATE a literature note at:\n      01 - NOTES/permanent/lit-[source-topic].md\n      \n      Literature note format:\n      ---\n      type: literature\n      source: [SOURCE NAME]\n      date_read: [TODAY]\n      topic: [TOPIC]\n      ---\n      \n      # [SOURCE TITLE]\n      \n      ## Core Argument\n      [What this source claims in my own words]\n      \n      ## Key Evidence  \n      [Specific evidence worth remembering]\n      \n      ## My Reaction\n      [What I agree with, doubt, or want to explore]\n      \n      ## Connections\n      [[RELATED PERMANENT NOTE 1]]\n      [[RELATED PERMANENT NOTE 2]]\n      \n      ## Ideas to Develop\n      [Specific ideas worth turning into \n      their own permanent notes]\n   \n   e. For each idea flagged in Ideas to Develop:\n      Create a new permanent note at\n      01 - NOTES/permanent/[concept-name].md\n   \n   f. Move original inbox item to archive\n\n4. Generate conversion summary:\n   [N] literature notes created\n   [N] permanent notes created\n   [N] connections to existing notes\n\n5. Save summary to:\n   04 - HERMES-OUTPUTS/analyses/[DATE]-research-converted.md\n\n## Quality Standard\nEvery literature note must produce at least \none new permanent note. If the research \nproduced no new ideas worth capturing as \npermanent notes the research was not worth \nprocessing in the first place.\n\nSkill 7: The Thinking Partner\n\nThis skill activates Claude as an active thinking partner that reads your recent notes and asks questions, surfaces contradictions, and pushes your thinking forward.\n\nCreate skills/thinking-partner.md:\n\n# thinking-partner\n\n## Purpose\nActivate Claude as an active thinking partner \nthat reads my recent vault activity and \nengages with my ideas rather than just \nstoring or summarizing them.\n\n## Trigger\nManual: \"Thinking partner session\" or \n\"Challenge my thinking on [TOPIC]\"\n\n## Process\n1. Read CLAUDE.md for current intellectual context.\n\n2. Read all permanent notes modified in \n   the past 14 days.\n\n3. Read recent daily notes from the past 7 days.\n\n4. Claude identifies:\n\n   INTERESTING TENSIONS:\n   Notes where I appear to hold positions \n   that are in tension with each other.\n   Not contradictions necessarily. \n   Tensions worth examining.\n   \n   UNDERDEVELOPED IDEAS:\n   Permanent notes that make a claim without \n   fully developing the reasoning or evidence.\n   \n   MISSING CONNECTIONS:\n   Ideas in recent notes that connect to \n   older permanent notes in non-obvious ways.\n   \n   OPEN QUESTIONS:\n   Questions raised in recent notes that \n   have not been answered anywhere in the vault.\n\n5. Claude engages with each finding:\n   Not just naming them. Pushing on them.\n   \n   For TENSIONS: \"You wrote X in [NOTE A] \n   and Y in [NOTE B]. These seem to pull \n   in different directions. Which position \n   do you actually hold and why?\"\n   \n   For UNDERDEVELOPED IDEAS: \"The note on \n   [TOPIC] makes the claim that [CLAIM]. \n   What is the strongest argument against \n   this? Does your vault contain evidence \n   that challenges it?\"\n   \n   For OPEN QUESTIONS: \"You asked [QUESTION] \n   in your note from [DATE]. You have not \n   answered it anywhere. Is this worth \n   investigating or has it become irrelevant?\"\n\n6. Save the session to:\n   04 - HERMES-OUTPUTS/analyses/[DATE]-thinking-partner.md\n\n## Tone\nThis skill should feel like talking to a \nsmart colleague who has read everything in \nyour vault and is genuinely engaging with \nyour ideas. Not validating. Not summarizing. \nPushing. Questioning. Connecting.\n\n## The Daily Experience\n\nWith the full system running the daily experience of your Obsidian vault changes completely.\n\n6:00 AM: Hermes reads your vault and generates the morning brief. You wake up to a note in HERMES-OUTPUTS/briefings/ that knows your project status, your open loops, your current priorities, and the external developments relevant to your work. Five minutes to read. You know exactly what matters before you see a single notification.\n\nDuring the day: You capture everything to 00 - INBOX. Ideas, tasks, references, meeting notes. No filing decisions. No thinking about where things go. Just capture.\n\n8:00 PM: Hermes processes your inbox automatically. Every capture gets filed, linked, and connected to your existing knowledge. Your INBOX empties itself. Nothing sits unprocessed.\n\nMonday 7:00 AM: Project health report waiting in HERMES-OUTPUTS/reviews. Every active project assessed. Stalled projects flagged. Next actions surfaced. Your week starts with a complete picture of where everything stands.\n\nSunday 5:00 PM: Connection report lands in HERMES-OUTPUTS. Links between recent notes and older permanent notes that you did not consciously make surface automatically. Your knowledge network gets denser every week without additional effort.\n\nSunday 7:00 PM: Weekly synthesis ready. The full week's vault activity synthesized into a document that shows what moved, what did not, what patterns emerged, and what next week's priorities should be. Your CLAUDE.md updates automatically with next week's top three priorities.\n\n## What Happens After 90 Days\n\nThe power of this combination is not visible on day one.\n\nIt becomes visible over time as two separate compounding effects interact.\n\nObsidian compounds because your permanent notes accumulate connections over months. A note written in January gets linked to notes written in March, April, and May. The network gets denser. The connection finder surfaces increasingly non-obvious relationships. The vault at month three is qualitatively richer than the vault at month one.\n\nHermes compounds because memory accumulates across every session. The morning brief at day ninety has read three months of your daily notes and knows your patterns. The weekly synthesis at week twelve has twelve weeks of previous syntheses to draw on. The project health monitor at month three knows how your projects typically stall and surfaces that pattern before it becomes a problem.\n\nThe interaction compounds because the more Obsidian knows the better Hermes can reason. The more Hermes produces the richer the vault becomes. The richer the vault the better the next Hermes output.\n\nBy month three you have a system that no longer feels like a tool you use.\n\nIt feels like a thinking partner who has been working alongside you for months, knows everything you have written, and surfaces the connections and insights your own working memory cannot hold.\n\nThat is the system this combination is capable of producing.\n\nBuild it this weekend.\n\nThe compounding starts from the first skill that reads your vault.\n\nFollow @cyrilXBT for every Obsidian system, Hermes Agent skill, and autonomous operation architecture that makes this combination run at its highest level."}}