The Agent Skills Directory

[COMMAND_EXECUTION]: The skill operates by executing a local binary named alera with various subcommands (capabilities, list-apps, get-app-state, click, set-value). This grants the agent direct control over the local desktop environment via the CLI.- [DATA_EXPOSURE]: The get-app-state command retrieves the accessibility tree and screenshots of desktop windows. This process can capture and expose sensitive information visible in open applications, such as personal details, documents, or private communications, to the agent's context.- [INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes content from untrusted external sources (desktop application windows).
Ingestion points: The alera computer get-app-state command reads the accessibility tree which contains the textual content of window elements.
Boundary markers: The documentation does not specify technical delimiters or boundary markers used in the tool's output to distinguish application data from instructions.
Capability inventory: The agent has the ability to perform actions such as click, set-value, and perform-secondary-action, allowing it to execute tasks within applications based on potentially malicious content read from those applications.
Sanitization: There is no mention of sanitization, filtering, or validation of the content retrieved from the accessibility tree before it is interpreted by the agent.

alera-computer-use