Relative link to script blocks publishing by triggering Confluence page lookup

[juancarlosjr97]

Description

A Markdown file containing a relative link to a shell script (for example, [this script](./scripts/setupServiceKey.sh)) causes mark to process the non-markdown file as a candidate Confluence page, attempting metadata extraction and a Confluence page API request. This results in a build failure when the API cannot resolve the script file as a Confluence page.

Reproduction

Directory structure:

test-confluence-syndication/
├── SERVICE_KEYS.md
└── scripts/
   └── setupServiceKey.sh

SERVICE_KEYS.md content:

# SERVICE_KEYS Testing

This is a test document with a relative link to [this script](./scripts/setupServiceKey.sh).

scripts/setupServiceKey.sh content:

#!/usr/bin/env bash
echo "test"

Run mark (v15.3.0):

docker run --rm -it -v "$PWD:/work" -w /work \
 kovetskiy/mark@sha256:5946929029b861ca89a46828292273a19d5694c833a4c07b5bd4c33660ea19b2 \
 mark --log-level TRACE \
   -b "https://confluence.example.com/wiki" \
   -u "user@example.com" \
   -p "YOUR_TOKEN" \
   --title-from-h1 \
   --space "YOUR_SPACE" \
   --parents "Parent Page" \
   -f "./SERVICE_KEYS.md"

Observed Behaviour

TRACE found a relative link: full=./scripts/setupServiceKey.sh filename=./scripts/setupServiceKey.sh hash=
TRACE filepath: scripts/setupServiceKey.sh
TRACE extracted metadata: space=~5c66e71d9f5b482e28d63e77 title=/usr/bin/env bash
TRACE rest: GET /wiki/rest/api/content/?expand=ancestors%2Cversion&spaceKey=~5c66e71d9f5b482e28d63e77&title=%2Fusr%2Fbin%2Fenv+bash&type=page
FATAL unable to resolve relative links
└─ resolve link: "./scripts/setupServiceKey.sh"
   └─ find confluence page: scripts/setupServiceKey.sh / ~5c66e71d9f5b482e28d63e77 / /usr/bin/env bash
      └─ api: find page
         └─ the Confluence API returned unexpected status: 403 Forbidden

Root Cause

Mark treats the shell script as a markdown file:

1. Detects the relative link ./scripts/setupServiceKey.sh
2. Attempts to extract metadata from the script file
3. Uses the shebang line #!/usr/bin/env bash as the page "title"
4. Tries to find a Confluence page with the title /usr/bin/env bash
5. Fails with 403 error

Expected Behaviour

Relative links to non-markdown files (e.g., .sh, .py, .txt) should not be treated as Confluence pages and should not block publishing.

Environment

• Mark version: v15.3.0
• Docker digest: sha256:5946929029b861ca89a46828292273a19d5694c833a4c07b5bd4c33660ea19b2
• Confluence: Cloud (A disposable Confluence created only to test this issue)

[juancarlosjr97]

I think a solution to this problem will be to add a file extension check before attempting metadata extraction in page/link.go