The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the retrieval of messages from Azure Storage Queues, which constitutes an external ingestion point for untrusted data. If the content of these messages contains malicious instructions, they could potentially influence the agent's logic or behavior.
Ingestion points: Untrusted data enters the agent's context through methods like receiveMessages() and peekMessages() as shown in references/detailed-guide.md.
Boundary markers: The provided implementation examples do not include boundary markers or explicit instructions for the agent to ignore command-like content within message bodies.
Capability inventory: The skill enables interactions with Azure infrastructure, including creating/deleting queues and sending/receiving data.
Sanitization: Examples show basic processing like JSON.parse(), but do not demonstrate sanitization or validation of the logical content retrieved from the queue.
[COMMAND_EXECUTION]: The installation guide includes standard package management commands for installing official SDKs from the public npm registry.
Evidence: npm install @azure/storage-queue @azure/identity in references/detailed-guide.md targeting official Microsoft packages.
[CREDENTIALS_UNSAFE]: The documentation references sensitive authentication methods involving account keys and connection strings. It correctly uses placeholders for these values, which is an expected and safe documentation practice.
Evidence: Placeholders like <account-key> and AccountName=... are utilized in references/detailed-guide.md rather than hardcoded credentials.

azure-storage-queue-ts