The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes work item summaries, labels, and epic data from Jira which is controlled by external users. This data is ingested into the agent's context and could contain instructions designed to manipulate the agent's output or actions.
Ingestion points: Work item details fetched via acli jira workitem search in workflows/summarize-sprint.md (Step 2).
Boundary markers: The instructions lack explicit delimiters or warnings to the model to ignore embedded instructions within the ticket data.
Capability inventory: The agent has the ability to execute shell commands (acli) and write files to the local system.
Sanitization: There is no process described for sanitizing or escaping the content of Jira issues before reporting.
[COMMAND_EXECUTION]: The skill workflow directs the agent to execute shell commands using the acli tool, with arguments constructed from user-supplied inputs such as Team ID and Board ID.
Evidence: Shell commands in workflows/summarize-sprint.md (Steps 1, 2, and 8) use placeholders like BOARD_ID and SPRINT_ID which are populated from user input.
[DYNAMIC_EXECUTION]: The skill generates a local markdown file based on a naming template that includes user-provided variables.
Evidence: Step 9 in workflows/summarize-sprint.md instructs the agent to write a file named sprint-report-{team}-{sprint}-{YYYY-MM-DD}.md to the current directory, which could lead to path traversal if the team or sprint variables contain directory manipulation characters.

rhdh-jira-sprint-report