The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of data from external Jira tickets through commands like jira-tool get and jira-tool search. If a ticket contains malicious instructions, they could influence the agent's subsequent actions.
Ingestion points: External ticket data is retrieved into the agent context in SKILL.md and QUICK_REFERENCE.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard instructions embedded within the Jira content.
Capability inventory: The skill uses subprocess.run and shell scripting to execute commands and process files.
Sanitization: No evidence of sanitization or escaping of the retrieved ticket content is present in the provided instructions.
[COMMAND_EXECUTION]: The skill's primary function is achieved through the execution of the jira-tool CLI and shell utilities like jq and cat via the command line.
[DATA_EXFILTRATION]: The skill provides commands for exporting Jira data to local files (jsonl, csv, json), which involves the transfer of data from a remote platform to the local filesystem for processing.
[PERSISTENCE_MECHANISMS]: The instructions suggest modifying shell profiles (~/.bashrc, ~/.zshrc) to store environment variables such as JIRA_API_TOKEN to ensure configuration persistence across shell sessions.
[DATA_EXPOSURE]: The troubleshooting section mentions echoing the JIRA_API_TOKEN, which can lead to sensitive credentials being recorded in shell history or visible in shared environments, though it includes a warning regarding shared terminals.

jira-builders