The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses a dedicated Chrome profile directory (defaulting to ~/.local/share/confluence-browser-fetch-chrome) to extract Atlassian session cookies. This sensitive information is used to authenticate requests to the Confluence REST API. While this is a high-privilege operation, it is a core part of the skill's documented functionality for handling SSO sessions.
[PROMPT_INJECTION]: The tool ingests content from local files and renders it into Confluence storage format, creating an indirect prompt injection surface. If the source files contain malicious instructions, they could be uploaded to Confluence and later affect other agents or users.
Ingestion points: Content is read from the file path specified in the --file argument within scripts/confluence-update.js.
Boundary markers: Absent. There are no delimiters or instructions provided to the agent to ignore potentially malicious content within the processed files.
Capability inventory: The skill is capable of performing network write operations (PUT and POST) to Atlassian Cloud endpoints as seen in scripts/confluence-update.js.
Sanitization: The scripts/lib.js file provides escapeHtml for Markdown rendering, but the tool allows for the direct ingestion of raw Storage XHTML content without additional sanitization.
[EXTERNAL_DOWNLOADS]: The script scripts/confluence-update.js requires a local module named atlassian-browser.js which is not included in the provided file list. Additionally, the documentation in references/distribution.md mentions the NPM package @aholbreich/agent-skills.

confluence-update