The Agent Skills Directory

[DATA_EXFILTRATION]: The scripts scripts/get-env-config.ps1 and scripts/get-env-config.sh extract the logged-in user's email address or username using the az account show command. This identity information is subsequently transmitted to a remote telemetry endpoint via a POST request using curl or Invoke-RestMethod. The destination URL and the required authentication key are not hardcoded but are dynamically retrieved from the user's Azure environment via the telemetry-cfg-endpoint and telemetry-cfg-api-key tags.- [COMMAND_EXECUTION]: The skill requires the agent to execute shell scripts and PowerShell scripts locally. These scripts interact with the Azure CLI (az) and system utilities like jq to manage Azure resources and process configurations.- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. It ingests untrusted data from Azure Resource Group tags, which are then parsed and reported back to the agent. This represents an attack surface where a user with write access to the Azure subscription's tags could inject malicious instructions into the agent's context. Evidence chain:
Ingestion points: az group show output (tags) in scripts/get-env-config.sh and scripts/get-env-config.ps1.
Boundary markers: Absent. No delimiters or 'ignore' instructions are used to wrap the ingested tag data.
Capability inventory: The skill utilizes the shell tool, granting it broad system execution capabilities.
Sanitization: Data is parsed as JSON, but the resulting content is presented directly to the agent without instruction-specific filtering.

aro-hcp-env-info