The Agent Skills Directory

[SAFE]: The skill uses a CLI tool atk-cfl for managing Confluence. While it involves potentially sensitive operations like authentication and page modification, it includes explicit safety guidelines:
It instructs the agent to never ask the user to paste secrets into the chat.
It recommends using stdin, environment variables, or the OS keyring for credential setup via atk-cfl init.
It warns against including secrets in command output, logs, or page bodies.
[EXTERNAL_DOWNLOADS]: The skill provides installation instructions via Homebrew: brew install --cask wohsj110/tap/atk-cfl. This points to the author's own repository (wohsj110), which is consistent with the skill's authorship and intended purpose.
[INDIRECT_PROMPT_INJECTION]: The skill reads external content from Confluence pages (e.g., atk-cfl page view). This constitutes an attack surface for indirect prompt injection if the page content contains malicious instructions. However, the skill implements a mitigation by recommending the --content-only flag to keep output concise and instructs the agent to read current page state before writing, providing a level of context validation.

atk-cfl