cascade-load-linked-resource-content

name	cascade-load-linked-resource-content
description	Sub-skill: Fetch raw content for one or more URLs via the extract-linked-resources MCP tool. Supports Jira issues, Confluence pages, Google Docs, and Google Sheets. Saves content to .temp/cascade/context/ and appends newly discovered links to to-load.md. Used as a building block by parent skills like generate-questions and write-jira-story.

Load Content

Fetch raw content for a set of URLs using Cascade MCP tools. Save content locally and discover new links for iterative loading.

When to Use

This is a sub-skill — called by parent skills (generate-questions, write-jira-story), not directly by users. Use when the parent skill needs to gather raw content from one or more URLs before analysis.

Prerequisites

• Cascade MCP server connected (tools available)
• URLs to load (from extract-linked-resources output or from to-load.md)

Procedure

1. Read the loading manifest

Check if .temp/cascade/context/to-load.md exists. If it does, read it to find URLs marked as [ ] (not yet loaded).

If the file doesn't exist, create it from the URLs provided by the parent skill:

# Links to Load

## Unloaded
- [ ] https://mycompany.atlassian.net/browse/PROJ-123
- [ ] https://docs.google.com/document/d/abc123/edit

## Loaded
(none yet)

2. Fetch content for each unloaded URL

For each [ ] URL, call the MCP tool extract-linked-resources with the URL:

extract-linked-resources({ url: "https://mycompany.atlassian.net/browse/PROJ-123" })

This returns markdown with YAML frontmatter containing:

• The document content (description, page body, etc.) as the markdown body
• discoveredLinks in frontmatter (categorized: figma, confluence, jira, googleDocs)
• For Jira: relationship on each link (parent, blocks, relates-to, etc.)
• For Jira: hasMoreComments / commentsStartAt for comment pagination

Figma URLs: If a Figma URL is passed, the tool returns a message to use figma-batch-zip instead. Figma URLs should NOT be loaded by this skill.

3. Save fetched content

Save the returned markdown+frontmatter directly as a file. The response is ready to write as-is:

.temp/cascade/context/
├── to-load.md                          ← loading manifest
├── jira-PROJ-123.md                    ← Jira issue (saved directly from tool response)
├── jira-PROJ-124.md                    ← linked Jira issue
├── confluence-page-title.md            ← Confluence page
├── gdoc-document-title.md              ← Google Doc
└── gsheet-spreadsheet-title.md         ← Google Spreadsheet

File naming: Use the source type prefix + a slugified identifier:

• Jira: jira-{issueKey}.md
• Confluence: confluence-{slugified-page-title}.md
• Google Doc: gdoc-{slugified-doc-title}.md
• Google Sheet: gsheet-{slugified-title}.md

4. Parse discovered links from frontmatter

Read the discoveredLinks YAML frontmatter from each saved file. Add any new URLs that aren’t already in to-load.md.

5. Update the loading manifest

Mark loaded URLs as [x] and append any newly discovered URLs as [ ]:

# Links to Load

## Unloaded
- [ ] https://mycompany.atlassian.net/wiki/spaces/TEAM/pages/12345
- [ ] https://www.figma.com/design/abc123/Designs

## Loaded
- [x] https://mycompany.atlassian.net/browse/PROJ-123
- [x] https://docs.google.com/document/d/abc123/edit

## Figma (handled separately)
- https://www.figma.com/design/abc123/Designs?node-id=0-1

Important: Figma URLs should be listed in a separate "Figma" section — they are NOT loaded by this skill. The parent skill handles Figma loading via figma-batch-zip (or figma-batch-cache in cloud environments).

6. Return to parent skill

Report back to the parent skill:

• How many URLs were loaded
• How many new URLs were discovered
• Whether there are remaining unloaded URLs in to-load.md

The parent skill decides whether to call load-linked-resource-content again (if new links were discovered) or proceed to analysis.

Important Notes

• Do NOT load Figma URLs — they require batch loading via figma-batch-zip + curl/unzip (or figma-batch-cache in cloud environments), which the parent skill handles
• Do NOT analyze or summarize content — that's the summarize-document-content sub-skill's job
• Deduplicate URLs — don't add a URL to to-load.md if it's already there (loaded or unloaded)
• Handle errors gracefully — if a tool call fails, mark the URL with [!] in the manifest and continue with other URLs