atlassian
Pass
Audited by Gen Agent Trust Hub on Jun 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute the
aclibinary for managing Atlassian work items and spaces. - [EXTERNAL_DOWNLOADS]: Instructions include downloading and installing the Atlassian CLI via Homebrew (
atlassian/tap/acli), which is a well-known third-party service. - [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) when processing data from external sources.
- Ingestion points: Untrusted data is ingested from Jira issue descriptions, comments, and Confluence page bodies via
acli jira workitem viewandacli confluence page view. - Boundary markers: There are no explicit boundary markers or instructions to the model to ignore embedded commands within the fetched Atlassian content.
- Capability inventory: The skill possesses significant capabilities, including the ability to create, edit, transition, and delete Jira work items, as well as manage Confluence spaces.
- Sanitization: The skill processes raw XHTML and JSON output from the CLI without sanitization or filtering, allowing potentially malicious instructions in ticket descriptions to influence agent behavior.
Audit Metadata