The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct the agent to fetch documentation and sample code from external URLs hosted on cdn.marmot-cloud.com. This domain is not recognized as a trusted source within the provided safety scope.
[COMMAND_EXECUTION]: The skill explicitly instructs the agent to use shell commands (curl -sL) to retrieve dynamic documentation. Executing network commands to fetch content that the agent then processes increases the risk of remote content manipulation.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted markdown data from external URLs which could contain malicious instructions.
Ingestion points: External markdown files are fetched via curl and read from cdn.marmot-cloud.com (e.g., product-decision.md, select-sdk.md).
Boundary markers: None identified to distinguish instructions from fetched data.
Capability inventory: The agent is empowered to generate code and provide integration logic based on this external data.
Sanitization: No validation or sanitization of the remote content is mentioned before the agent processes it.

antom-integration