Skip to main content
Skills are system instruction files that tell your AI agent exactly how to work with xysq. Install a skill once and your agent knows how to authenticate, what to capture, when to recall, and how to use tags — without you explaining it every session.

​
How it works

Skills come in two tiers: Baseline skill — available immediately, no memories required. A hand-crafted set of instructions covering the core protocol: session start, aggressive capture, surgical recall, knowledge handling, and consent rules. Works from the moment you install it. Personalised skill — unlocked once you have more than 5 memories. xysq reflects on your memory bank, pulls answers to five curated questions about how you work, and injects a personalised section into the skill. Your agent starts each session already knowing your working style, active projects, tool preferences, and agent behaviour notes.

​
Installing your skill

​
Step 1 — Download from the dashboard

Go to app.xysq.ai and open the Dashboard. The Get Started panel has a Download Skill button. Click it — a dialog opens with download options for every supported platform.

​
Step 2 — Choose your platform

PlatformFileWhere to install
Claude Code.md~/.claude/skills/xysq.md
Claude Desktop.mdPaste into Project Instructions
Cursor.cursorrulesProject root as .cursorrules
Windsurf.windsurfrulesProject root as .windsurfrules
ChatGPT.mdPaste into Custom GPT instructions
Generic.mdAdd to your agent’s system prompt
Download the file for your platform and follow the install path above.

​
Step 3 — Verify it’s active

The next time your agent starts a session, it should automatically call authenticate() and memory_reflect() without you asking. That’s the skill working.

​
Getting a personalised skill

Once you have more than 5 memories stored, the Get Started panel will prompt you to generate a personalised skill. Click Download personalised skill — xysq runs five reflect queries against your memory bank and injects the results into the skill file as a personalised section. The personalised section covers:
  • About you — your primary areas of work and expertise
  • Working preferences — how you like to work with agents
  • Active context — current projects and focus areas
  • Agent behaviour notes — feedback you’ve given about AI interactions
  • Key resources — your most important tools and references
Download the updated file and replace your existing skill installation.

​
Keeping your skill current

Your skill evolves as your memory bank grows. xysq checks whether your memories have changed meaningfully since the last generation — if they have, a new version is written. Client-side daily refresh — the dashboard silently checks whether your personalised skill is more than 24 hours old each time you open it. If it is, it regenerates in the background and updates the version number. You’ll see the new version available to re-download. Via skill_sync() — if you’re using Claude Code or Cursor with an active MCP connection, your agent can fetch the latest skill directly: 
skill_sync()
The tool returns the current skill content and the recommended install path. For Claude Code, the agent can write the file directly. For other platforms it will return the content for you to paste.

​
Via your agent (MCP)

Once connected to xysq via MCP, your agent can manage its own skill installation: 
Sync my xysq skill

Update my skill file with the latest version
Claude Code will write ~/.claude/skills/xysq.md directly. Cursor and Windsurf will return the content for you to save into the project root.

​
The north star

Every skill is written around a single principle: less tokens = same or better output. Agents that use xysq effectively don’t need you to re-explain your context at the start of every session. They don’t ask what project you’re working on. They don’t forget your preferences. They start informed. The skill teaches your agent to capture aggressively — every decision, correction, preference, and fact — so that future sessions cost fewer tokens to reach the same quality of output. The memory you build today is tokens saved tomorrow.

​
What the skill instructs

Session start protocol
  1. Call authenticate() — establishes identity and checks memory health
  2. Call memory_reflect("What should I know about this user to assist them today?") — loads your context as working memory without asking you to re-explain it
Capture rules
What happenedSignificanceAction
You made a decisionhighRetained immediately with decision tag
You gave a correctionhighRetained immediately with correction tag
You expressed a preferencenormalRetained with preference tag
You stated a fact about your contextnormalRetained with fact tag
You referenced an external resource—Added via knowledge_add()
Transactional reply (“yes”, “ok”)—Skipped
Recall hierarchy (cheapest → most expensive)
  1. memory_reflect(query) — synthesis question, always tried first
  2. memory_recall(budget="low") — targeted fact lookup
  3. memory_recall(budget="mid") — broader context pull, used sparingly
  4. memory_recall(budget="high") — last resort, only if you ask for everything

​
Skill versions

Every time a new personalised skill is generated, the version number increments. All versions are stored in GCS — nothing is deleted. The current version is always what your agent gets via skill_sync(). You can see the current version and when it was last generated from the Dashboard.