The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of the ae-cli command-line utility to perform operations on audiences and tags. Commands are constructed dynamically using project IDs, names, and JSON-formatted definitions provided by the user or derived from previous tool outputs.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface with the following evidence chain:
Ingestion points: Untrusted data enters the agent context via ae-cli analysis_audience +list_cluster_members (in references/list-cluster-members.md) and +list_tag_members (in references/list-tag-members.md), which retrieve user-defined property values from the audience database.
Boundary markers: There are no instructions or structural markers, such as delimiters or explicit warnings to ignore embedded commands, present to prevent the agent from being influenced by instructions hidden within the retrieved user properties.
Capability inventory: The skill possesses significant capabilities, including creating and updating clusters/tags and triggering recomputations through ae-cli subprocess commands across multiple reference files.
Sanitization: The skill documentation and command structures lack evidence of logic to sanitize, escape, or validate the content of external user properties before they are processed or used to construct subsequent command definitions.

te-audience