The Agent Skills Directory

[COMMAND_EXECUTION]: The helper script scripts/_common.py executes a local Python script scripts/check_auth_dependency.py using subprocess.run. This is used as a benign internal check to verify that a required dependency skill is installed in the current environment.
[DATA_EXFILTRATION]: The skill sends advertisement management data and user authentication tokens to vendor-owned domains, including tool-gateway.linkfox.com, api.linkfox.com, and agent-api.linkfox.com. These are legitimate endpoints necessary for the operation of the Linkfox service.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external APIs (Amazon Ads via proxy), which represents an attack surface for indirect prompt injection.
Ingestion points: API responses containing advertisement metadata are handled in scripts/_common.py and printed to the agent's context.
Boundary markers: No explicit delimiters or instructions are implemented to prevent the agent from potentially obeying instructions embedded in ad fields.
Capability inventory: The skill can perform network operations, write to local files, and execute its own internal scripts.
Sanitization: The skill relies on standard JSON parsing; however, the content of the data fields is not specifically sanitized against malicious text-based instructions.
[EXTERNAL_DOWNLOADS]: The scripts/onboarding.py utility suggests that users install standard Python libraries (requests, qrcode, pillow) if they are missing. These are common dependencies, and the skill does not perform any automated or unverified downloads.
[CREDENTIALS_UNSAFE]: The skill accesses API keys from environment variables (LINKFOX_AGENT_API_KEY) for service authentication. It follows secure practices by instructing users to manage these secrets through their shell configuration rather than hardcoding them in the skill content.

linkfox-amazon-ads-manager