← Back to Docs

Skill: Update Memory

Skill: Update Memory

The update_memory skill teaches Claude or Codex how to retrieve an existing Memory from Context Link, merge in new information from the current conversation, and save it back. It's a GET-then-POST workflow that keeps your saved Memories current without starting from scratch.

This is how Memories become living documents. Your /brand-voice evolves as your tone sharpens. Your /roadmap shifts as priorities change. Your /keyword-tracker grows as you publish new content. Instead of replacing the whole thing, Claude fetches what's already there, weaves in the new details, and saves the updated version.

When to use this skill

  • A project spec has evolved and you want to merge in the latest decisions without rewriting everything
  • You've refined your brand voice or messaging and want to update the saved version
  • You want to append new findings to an existing research summary or competitor analysis
  • You're iterating on a document over multiple sessions and want each version stored under the same slug

For saving something new (not updating an existing Memory), see the Save Memory skill. For retrieving context without modifying it, see the Get Context skill.

How it works

When triggered, the skill:

  1. GETs the existing content from your Context Link using the slug (e.g. /brand-voice)
  2. Merges the old content with new information from the conversation — deduplicating, reorganising, and keeping it concise
  3. POSTs the updated content back to the same slug

This creates a new version under the same slug. Old versions are preserved in the background, but retrieval always returns the latest. Two requests total — one GET, one POST.

The skill file

Below is the full skill definition. Replace YOUR_CONTEXT_LINK with your personal Context Link URL (e.g. yourname.context-link.ai). If you use a PIN, append ?p=YOUR_PIN to the URL.

You can download a pre-populated version from the Installation page in your dashboard — no manual editing needed.


---
name: update-memory
description: Retrieve saved content from Context Link, update it with current context, and save it back
---

**What is Context Link?** Context Link is an external service that indexes connected sources (websites, Google Drive, Notion) and memories into a searchable knowledge base. It provides semantic search and memory storage via a simple URL: `subdomain.context-link.ai/query?p=optional_pincode`. If you don't know the user's Context Link URL, ask them for it.

---

### Update Content on Context Link

Retrieve existing content, merge in new information, and save it back. Two requests total — one GET, one POST.

**Workflow:**

1. **Print this message:** `🔗 Updating memory on Context Link → {SLUG}` — Never print the actual Context Link URL, as it contains a private 'pin' or 'p' URL param.

2. **GET the existing content.**

```bash
curl -s "YOUR_CONTEXT_LINK/{SLUG}"
```

Returns HTML. Extract the text content from inside the `<body>` tag — ignore HTML boilerplate.

3. **Merge and rewrite.** Combine existing content with new information from the conversation. Deduplicate, reorganize if needed, keep it concise. Output as clean markdown.

4. **POST the updated content back to the same slug.**

```bash
curl -s -X POST "YOUR_CONTEXT_LINK/{SLUG}" \
  -H "Content-Type: text/plain" \
  -d 'Your updated markdown here'
```

The body is raw text/markdown — not JSON. The server handles chunking internally.

**Success response:** `{"message": "Saved", "namespace": "{SLUG}"}` with HTTP 201.

**Rules:**
- **Keep the body under 100KB.** If the merged content is too long, summarize or condense it before POSTing.
- Two requests max: one GET, one POST. Nothing else.
- Do NOT ask the user what to update unless genuinely unclear — infer from conversation.
- This creates a new version under the same slug (old versions still exist, slug points to latest).
- After saving, confirm briefly: "Updated `{SLUG}` on Context Link."
- If the request is blocked, ask the user to add `*.context-link.ai` to Claude's **Settings → Capabilities → Domain Allowlist** (or select "All domains"), then retry.

Installation

Claude (Chat, Cowork, and Code)

Skills work with Claude Chat, Claude Cowork, and Claude Code. (If you're using Claude Cowork, you can also use the Context Link Plugin to install all skills at once.)

To install this skill:

  1. In Claude Chat or Claude Cowork, click Customize in the top right
  2. Click Skills, then +, then Upload a skill
  3. Upload the skill file (download from Installation)
  4. Allow network access (see below)

For Claude Code, place the skill file in your project's .claude/skills/ directory.


Allow network access

For Claude to reach Context Link, you need to allow the domain in Claude's settings:

  1. In Claude, go to Settings → Capabilities → Domain Allowlist
  2. Add *.context-link.ai to the allowlist (or select "All domains")

Claude domain allowlist settings

Once installed, tell Claude to "update [topic] on Context Link" and it handles the fetch-merge-save cycle.

OpenAI Codex

  1. Download the skill files from Installation
  2. Unzip and place the update_memory folder into ~/.agents/skills/ (or your repo's .agents/skills/ directory)
  3. Codex auto-detects the skill

See the Codex skills documentation for more on skill file locations and priority.

Example usage

Updating a brand voice doc after a positioning shift:

You: We've decided to drop the "revolutionary" language and lean into the pragmatic angle more. Update brand-voice on Context Link.

AI: GETs the existing /brand-voice Memory, merges in the new tone direction, removes the "revolutionary" references, POSTs it back. "Updated brand-voice on Context Link."

Adding new milestones to a roadmap:

You: Update project-roadmap on Context Link with these new milestones we just discussed.

AI: Fetches the current roadmap, weaves in the new milestones, reorganises by priority, saves. The next time anyone on the team asks their AI for context on the roadmap, they get the latest version.

Evolving a keyword tracker after publishing new content:

You: We just published the AI knowledge base article. Update keyword-tracker with the new published status and URL.

AI: GETs the tracker, marks the keyword as published, adds the URL, POSTs the updated version.

The skill infers what needs updating from the conversation context, so you don't need to repeat everything — just point it at the right slug and describe what changed.

How versioning works

  • Each update creates a new version under the same slug
  • Old versions are preserved in the background
  • Retrieving by slug (via the Get Context skill, a direct link, the ChatGPT connector, or the API) always returns the latest version
  • This means you can update freely without worrying about losing previous content

The bigger picture: READ and WRITE

Context Link has two core capabilities:

  • READ (Get Context) — search your connected sources (Notion, Google Docs, websites) and your saved Memories by meaning, and return the most relevant snippets
  • WRITE (Save + Update) — create and maintain living documents that your AI can build on over time

The update skill is where these two sides meet. You're reading what's already saved, improving it with new information, and writing it back. Over time, your Memories become a curated, evolving knowledge base that any AI tool can draw from — not just Claude, but ChatGPT, Copilot, Gemini, or anything that can reach your Context Link URL.



Don't have a Context Link account yet? Sign up and connect your first source — start building living documents your AI can maintain and improve over time.

← Back to Docs