The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external Atlassian services, such as Jira issue descriptions, comments, and Confluence page content. Because this data is user-controlled, it represents a surface for indirect prompt injection attacks where malicious instructions could be embedded in tickets to influence the agent's behavior.
Ingestion points: Commands such as acli jira workitem search, acli jira workitem view, and acli confluence page view read content into the agent's context.
Capabilities: The skill has broad capabilities including execution of acli commands and curl requests to the Atlassian API.
Sanitization: Includes a dedicated ADF (Atlassian Document Format) converter and validator in scripts/md-to-adf.ts to ensure structural integrity of data sent to Jira, which provides some protection against malformed payloads but not natural language injection.
[EXTERNAL_DOWNLOADS]: The documentation and quick start guides provide instructions for downloading the acli binary from acli.atlassian.com. This is the official distribution domain for the Atlassian CLI tool and is considered a well-known service associated with the tool's vendor.
[COMMAND_EXECUTION]: The skill functions by executing the acli CLI tool and curl commands via the Bash tool. It also includes utility scripts (scripts/md-to-adf.ts, scripts/jira-attach-media.ts) executed via the bun runtime for handling rich text conversion and media attachments.

acli