The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructions provide commands to extract API keys from a sensitive local file path at ~/.gooseworks/credentials.json. Accessing credential files is a high-risk operation that can lead to local data exposure.
[EXTERNAL_DOWNLOADS]: The documentation suggests running npx gooseworks login, which downloads and executes the gooseworks package from the npm registry. This involves executing unversioned code from an external repository at runtime.
[COMMAND_EXECUTION]: The skill utilizes shell commands (python3 -c) to parse local JSON configuration and curl to transmit data to external APIs. While functional, these operations rely on the agent's ability to execute arbitrary commands, which is a significant capability.
[PROMPT_INJECTION]: The skill processes untrusted external content from Amazon search results, creating an attack surface for indirect prompt injection.
Ingestion points: Untrusted product data (titles, descriptions, delivery info) enters the agent's context through the organic_results array returned by the API.
Boundary markers: The prompt instructions do not include clear delimiters or safety warnings to help the agent distinguish between its core instructions and the external data being processed.
Capability inventory: The skill has access to network tools (curl) and shell execution (python3), which could be misused if a malicious instruction from a product listing is executed by the agent.
Sanitization: The skill does not define any sanitization or validation steps for the Amazon content before it is processed by the model.

amazon-search