The Agent Skills Directory

[COMMAND_EXECUTION]: The script _spapi_fulfillment_outbound_common.py invokes subprocess.run() to execute a companion local dependency script check_auth_dependency.py. This invocation uses a fixed array and does not pass input through a shell interpreter, making it resilient against command injection.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from Amazon SP-API endpoints and renders summaries to the agent context.
Ingestion points: External fulfillment and order tracking data is pulled from Amazon's API gateway via _linkfox_gateway.py.
Boundary markers: The system lacks explicit boundary wrappers or specific instructions for processing untrusted strings returned within customer tracking details or product descriptions.
Capability inventory: The skill includes data persistence capabilities (emit_result writes responses to local files) and standard shell executions for local dependency checks, but does not execute dynamic code (eval/exec) derived from API payloads.
Sanitization: Request inputs are rigorously validated before being sent out via _spapi_fulfillment_outbound_validators.py, but the fields parsed from API responses are not filtered or sanitized prior to output generation.

linkfox-amazon-store-fulfillment-outbound