The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/atlassian-browser.js executes local browser binaries using child_process.spawn. It searches for common Chromium-based browsers (Chrome, Chromium, Brave, Edge, Vivaldi) in the system PATH and standard installation directories.
[DATA_EXFILTRATION]: The module scripts/atlassian-browser.js programmatically extracts session cookies from a dedicated Chromium profile using the Network.getCookies method of the Chrome DevTools Protocol. These cookies are subsequently used to authenticate REST API requests to Atlassian domains.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from remote Confluence pages, which could potentially contain malicious instructions affecting downstream agent tasks.
Ingestion points: scripts/confluence-browser-fetch.js fetches page JSON, storage/view HTML, and browser-rendered HTML from user-specified Atlassian sites.
Boundary markers: The skill does not implement delimiters or 'ignore embedded instructions' warnings when saving the fetched content to the local file system.
Capability inventory: The skill has the ability to spawn shell processes (via scripts/atlassian-browser.js) and write arbitrary files to the local output directory (via scripts/confluence-browser-fetch.js).
Sanitization: While scripts/lib.js provides safeName to sanitize filenames against directory traversal, the actual content of the Confluence pages is stored in its raw, unsanitized form.

confluence-browser-fetch