The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted ticket content from external systems (Jira, GitHub) which could contain malicious instructions.
Ingestion points: Untrusted ticket data enters the agent context through the {FULL_TICKET_DESCRIPTION} placeholder in the Sub-Agent Prompt Template within SKILL.md.
Boundary markers: The skill uses Markdown headers (e.g., ## Ticket Details) to separate ticket content, but lacks explicit "ignore embedded instructions" warnings to prevent the agent from obeying commands found within tickets.
Capability inventory: The skill possesses significant capabilities including writing to ticket systems (addCommentToJiraIssue, gh issue comment), executing shell commands via acli and git, and performing deep codebase analysis via MCP tools.
Sanitization: No sanitization or filtering is performed on ticket content before it is interpolated into the sub-agent's prompt.
[COMMAND_EXECUTION]: The skill instructs the agent to execute various shell commands to perform its duties.
Evidence: The skill uses git remote get-url, git rev-parse, the GitHub CLI (gh), and the Atlassian CLI (acli) for ticket management and repository resolution.
Context: While these are standard developer tools, the use of acli involves passing variables like ticket keys and comment IDs directly to the shell, which requires the agent to be cautious of command injection if those fields are manipulated.

ticket-grooming