Bug: filter out external links from documentation

[walmsles]

Bug Description

Sometimes, a search will find a community resource that has an external link. The MCP fetch_docs tool will only load pages from the powertools documentation domain, so these will fail.

Steps To Reproduce

Hard to reproduce because of the LLM Search.

Expected Behavior

fetch_doc should filter out links that are external.

Actual Behavior

Tried to fetch the page but returned empty response (which is okay - but can bet better)

[walmsles]

Just thinking now on this one. Should look to allow domains from the community page so the MCP server is able to load community-linked pages. The fetch_docs tool should only load from allowed domains so that it cannot be used to load arbitrary pages or move out of the scope of powertools-mcp

[walmsles]

Idea: Gather all Links from doc pages and add them to an allow list. We don't want to allow loading of random pages but any page linked to from powertools docs should be allowed.

[walmsles]

Some Further Information / Context

The search index includes community pages for "I built this". The links for these pages are external to the powertools website. As a security precaution, the "fetch_doc" tool will only return content for pages that are on the powertools doc domain "docs.powertools.aws.dev".

This is less of a bug and more of an oversight in being careful about what the MCP server allows access to, with the assumption that all pages will be on the main powertools doc site domain.

Number of options:

1. Ignore pages that list external content.
2. Filter out external links from search results
3. Dynamically add detected external links as search results are sent back to the MCP client so if a fetch comes thorugh it will be allowed.

I currently lean towards one (1) as the direction to go. It is the simplest response for these when they do turn up and the idea of powertools is to help AI code - so going to external sites is not helpful within that scope.