The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from Jira issues, which are external and potentially user-controlled sources. \n
Ingestion points: The agent reads ticket details and comments via mcp__atlassian__getJiraIssue and jira issue view (SKILL.md, references/mcp.md). \n
Boundary markers: The skill instructions require the agent to show tool calls and obtain user approval before any modifications (SKILL.md). \n
Capability inventory: The skill can transition issues, update descriptions, and execute shell commands through the Jira CLI (SKILL.md, references/commands.md). \n
Sanitization: There is no evidence of sanitization for incoming Jira ticket data before it is processed by the agent. \n- [COMMAND_EXECUTION]: The skill instructions utilize subshell execution patterns, such as $(jira me), to dynamically inject user identity into commands (SKILL.md, references/commands.md). It also references a local script jira-link-issues for handling specific linking tasks (references/mcp.md). \n- [DYNAMIC_EXECUTION]: To bypass CLI limitations for multi-line text, the skill generates a temporary file in /tmp/jira_body.md and uses command-line interpolation to read its contents during issue creation (references/commands.md). \n- [EXTERNAL_DOWNLOADS]: The documentation recommends the installation of the jira-cli tool from a GitHub repository. This is presented as a standard prerequisite for the CLI backend functionality (README.md).

jira