The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes natural language queries from users and conversational responses from the Amazon Alexa API. This creates a standard surface where external data is ingested into the agent context. The skill has the capability to write results to the project directory and perform network requests to the vendor's gateway.
Ingestion points: User-provided prompts and url parameters defined in SKILL.md; conversational response data processed in scripts/amazon_alexa_search.py.
Boundary markers: The skill relies on standard conversational boundaries; no explicit delimiters or "ignore instructions" markers are enforced on the API responses.
Capability inventory: Network access via urllib to the LinkFox API gateway; file system write access to the linkfox/ directory within the workspace for session data and payment QR codes.
Sanitization: No specific filtering or sanitization of the Alexa API's conversational output is performed before presentation to the agent/user.
[EXTERNAL_DOWNLOADS]: The scripts/onboarding.py utility includes check-and-notify logic for the requests and qrcode Python libraries. If these dependencies are missing, the script provides standard installation instructions (pip install) to the user. These are well-known, legitimate packages.
[DATA_EXFILTRATION]: The skill transmits search prompts and session identifiers to tool-gateway.linkfox.com. The onboarding script transmits authentication data (phone numbers and verification codes) to api.linkfox.com and agent-api.linkfox.com. All network operations target official infrastructure belonging to the skill's author, 'linkfox-ai', and are necessary for the skill's primary functions (search, account registration, and credit management).
[COMMAND_EXECUTION]: The skill executes Python scripts to facilitate API communication and local data management. These operations are consistent with the documented purpose of providing a shopping assistant and handling user onboarding.

linkfox-amazon-alexa-search