The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a detailed reference for the acli command-line tool. It instructs the agent to construct and execute shell commands to interact with Jira Cloud and Atlassian organizations. The skill includes specific safety instructions for the agent, such as requiring confirmation for destructive operations and enumerating targets before bulk changes.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from Jira (e.g., issue summaries, descriptions, or JQL search results) and interpolates them into commands or the agent context.
Ingestion points: Untrusted content enters the agent context through search and view commands (e.g., acli jira workitem search, acli jira workitem view) documented in references/jira-workitem-commands.md and references/other-commands.md, as well as through user-supplied inputs for creation and editing.
Boundary markers: The skill does not provide explicit boundary markers or delimiters to isolate external content from instructions.
Capability inventory: The skill enables a wide range of capabilities via acli, including reading, creating, modifying, and permanently deleting Jira issues, projects, fields, and user accounts.
Sanitization: The skill includes mitigation instructions requiring the agent to obtain fresh user confirmation for destructive actions and to enumerate resolved targets before executing commands.

acli