The Agent Skills Directory

Command Execution: The skill provides a set of shell scripts (cql_search.sh, read_page.sh, write_page.sh) that use common system utilities such as curl and jq to interface with Confluence. This is used for the skill's primary purpose of managing wiki content.
External API Interaction: The skill communicates with Atlassian's Confluence Cloud endpoints. It uses standard HTTP Basic authentication with credentials expected to be managed by the host environment.
Indirect Prompt Injection Considerations: The documentation includes a prominent security note advising that content retrieved from the wiki should be treated as untrusted, as it may contain instructions authored by third parties. This is a well-documented consideration for agents processing external data.
Defensive Scripting Patterns: The included bash scripts implement several security measures, such as validating that API-provided pagination links are relative to prevent redirection and enforcing path-depth checks to ensure local file reads are confined to designated temporary directories.

confluence-api