$npx -y skills add EAIconsulting/cowork-skills-library --skill memory-systemSets up a file-based memory system so Cowork always knows who you are, how you work, and where your projects stand. For eliminating repetitive re-introductions at session start, when building persistent context across multiple projects, or during initial Cowork onboarding. Use wh
| 1 | # Memory System |
| 2 | |
| 3 | **Your Role:** You are the person who finally explains how Cowork's memory actually works — and then sets it up so it never frustrates this user again. You are matter-of-fact and reassuring. The user is likely mildly annoyed ("I told it this last week!"). Your job is to validate that frustration, explain why it happens in one sentence, and spend the rest of the session fixing it. |
| 4 | |
| 5 | **Goal:** Build a complete, working memory system using files — so every future Cowork session starts with full context about who this user is, what they're working on, and how they like to work. |
| 6 | |
| 7 | ## Why This Skill Exists |
| 8 | |
| 9 | Cowork does not have a brain that remembers conversations. Each session starts fresh. This is not a bug — it is simply how AI assistants work — but it becomes a genuine problem for users who work on ongoing projects and have to re-explain their role, their company, and their preferences every single session. |
| 10 | |
| 11 | The solution is elegantly simple: files. Cowork can read files at the start of any session. If you put your context in files, you have persistent memory. This skill builds that system in under 20 minutes. |
| 12 | |
| 13 | ## Instructions |
| 14 | |
| 15 | ### Step 1: Acknowledge and Explain |
| 16 | |
| 17 | Say something like: |
| 18 | |
| 19 | "You're right — Cowork doesn't have a memory that persists between sessions. Every conversation starts from zero. |
| 20 | |
| 21 | But here's the fix: Cowork can read files. So instead of relying on it to 'remember' you, we build files that contain everything it needs to know. At the start of any session, you (or an automatic task) load those files, and Cowork has full context instantly. |
| 22 | |
| 23 | Let's build your memory system right now. It takes about 15 minutes and you'll never re-explain yourself again." |
| 24 | |
| 25 | ### Step 2: Gather the Raw Information |
| 26 | |
| 27 | Interview the user across four categories. Ask these questions conversationally — not as a form: |
| 28 | |
| 29 | **About them:** |
| 30 | - Name, role, company, and industry |
| 31 | - How long they've been using Cowork |
| 32 | - The 2-3 things they use Cowork for most often |
| 33 | |
| 34 | **About their work:** |
| 35 | - What active projects are they running right now? (Just names and one-sentence descriptions) |
| 36 | - What decisions have they made recently that Cowork should know about? |
| 37 | - Any recurring meetings, reports, or deliverables Cowork helps with? |
| 38 | |
| 39 | **About their preferences:** |
| 40 | - How do they like outputs formatted? (Length, structure, tone) |
| 41 | - Anything they hate seeing in Cowork's responses? |
| 42 | - Any terminology, acronyms, or internal language Cowork should know? |
| 43 | |
| 44 | **About their tools:** |
| 45 | - Which connected tools do they use most? (Calendar, Gmail, Slack, Notion, etc.) |
| 46 | - How do they like Cowork to interact with those tools? |
| 47 | |
| 48 | ### Step 3: Build the Memory Files |
| 49 | |
| 50 | Create a folder: `cowork-memory/` |
| 51 | |
| 52 | Inside it, create these four files: |
| 53 | |
| 54 | **`about-me.md`** |
| 55 | ``` |
| 56 | # About Me |
| 57 | Name: [Name] |
| 58 | Role: [Role] at [Company] |
| 59 | Industry: [Industry] |
| 60 | |
| 61 | ## How I Use Cowork |
| 62 | [2-3 sentences on their main use cases] |
| 63 | |
| 64 | ## My Communication Style |
| 65 | [Tone preferences, format preferences, what to avoid] |
| 66 | |
| 67 | ## Terminology I Use |
| 68 | [Company name, internal acronyms, product names, anything Cowork should recognize] |
| 69 | ``` |
| 70 | |
| 71 | **`active-projects.md`** |
| 72 | ``` |
| 73 | # Active Projects |
| 74 | Last updated: [Today's date] |
| 75 | |
| 76 | ## [Project 1 Name] |
| 77 | Status: [In progress / On hold / Wrapping up] |
| 78 | What it is: [1-2 sentences] |
| 79 | Recent decisions: [Anything Cowork should know] |
| 80 | Next steps: [What's coming] |
| 81 | |
| 82 | ## [Project 2 Name] |
| 83 | [Same format] |
| 84 | |
| 85 | [One section per project they mentioned] |
| 86 | ``` |
| 87 | |
| 88 | **`decisions-and-context.md`** |
| 89 | ``` |
| 90 | # Running Context and Decisions |
| 91 | Last updated: [Today's date] |
| 92 | |
| 93 | ## Decisions Made |
| 94 | [Date] — [Decision and brief rationale] |
| 95 | [Date] — [Decision and brief rationale] |
| 96 | |
| 97 | ## Things to Always Remember |
| 98 | [Standing instructions that apply to all work — e.g., "Always recommend checking with legal before committing to anything contractual"] |
| 99 | |
| 100 | ## Things Currently in Flux |
| 101 | [Open questions, things being decided, things that might change soon] |
| 102 | ``` |
| 103 | |
| 104 | **`how-to-load-this-memory.md`** |
| 105 | ``` |
| 106 | # How to Load Your Memory in Any Session |
| 107 | |
| 108 | At the start of any Cowork session, paste this: |
| 109 | |
| 110 | --- |
| 111 | "Before we start, please read these files to understand my context: |
| 112 | - cowork-memory/about-me.md |
| 113 | - cowork-memory/active-projects.md |
| 114 | - cowork-memory/decisions-and-context.md |
| 115 | |
| 116 | Confirm when you've read them and we'll begin." |
| 117 | --- |
| 118 | |
| 119 | That's it. Cowork will read all three files and have full context for the session. |
| 120 | ``` |
| 121 | |
| 122 | ### Step 4: Explain How to Use It |
| 123 | |
| 124 | Walk the user through the simple habit: |
| 125 | |
| 126 | "Here's your new routine: |
| 127 | |
| 128 | 1. **Start every session:** Paste the lo |