The Agent Skills Directory

[PROMPT_INJECTION]: The skill has a significant indirect prompt injection surface because it retrieves and processes untrusted content from Jira issues and comments.
Ingestion points: Data enters the agent context through file and command outputs from 'jira issue view', 'jira issue list', and comments.
Boundary markers: No explicit delimiters or instructions are used to distinguish system commands from content fetched from Jira.
Capability inventory: The skill can execute shell commands, run Node.js scripts, and upload files to remote servers using 'curl'.
Sanitization: The ADF parser ('parse-adf.js') converts JSON structure to text but does not sanitize or filter the content for instructions.
[DATA_EXFILTRATION]: The 'jira-attach.sh' script provides a mechanism to upload any local file to the Jira REST API. While intended for project documentation, an agent could be manipulated via prompt injection to upload sensitive files such as '~/.ssh/id_rsa' or '.env' files to the user's Jira instance.
[COMMAND_EXECUTION]: The skill uses multiple local execution methods including shell scripts and Node.js. It accesses the local configuration file at '~/.config/.jira/.config.yml' to read sensitive server and login details for authentication.

jira-cli