The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves and analyzes product data (such as titles, bullet points, and descriptions) from Amazon via the ZooData API. Because this data is sourced from external sellers, it represents an attack surface for indirect prompt injection.
Ingestion points: Untrusted product content from api.zoodata.ai is processed by the agent in SKILL.md workflows.
Boundary markers: The skill instructions mandate the use of structured output and specific confidence labels (📊, 🔍, 💡) to separate external data from internal reasoning.
Capability inventory: The skill uses the zoodata.py script to perform network requests and manage local temporary result files.
Sanitization: The zoodata.py utility implements host-validation for credentials and uses standard library JSON parsing to handle external data safely.
[COMMAND_EXECUTION]: The bundled zoodata.py script uses subprocess.Popen to manage background maintenance tasks through self-invocation.
Evidence: In scripts/zoodata.py, the script invokes its own file path using sys.executable and os.path.abspath(__file__) to run a detached cleanup worker.
Context: This execution is used exclusively to clean up stale temporary result directories. The script verifies directory ownership and restricts permissions (0700) to ensure that these operations are isolated to the current user's private namespace.

amazon-pricing-command-center