The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's documentation instructs users to install and run the MCP server using npx -y atlascloud-mcp across various CLI tools (Claude, Gemini, etc.). It also provides a command to add the skill via npx skills add AtlasCloudAI/atlas-cloud-skills.
[INDIRECT_PROMPT_INJECTION]: The skill features a vulnerability surface where data from external sources is ingested into the agent context.
Ingestion points: Tools like atlas_transcribe_audio return text derived from external audio files, and atlas_chat retrieves content from various third-party LLMs.
Boundary markers: There are no instructions or patterns provided to delimit or treat the external data as untrusted.
Capability inventory: The skill includes file reading capabilities via atlas_upload_media and multiple network-facing tools for API interaction.
Sanitization: The skill does not describe any sanitization or validation logic for the content returned by the APIs.
[DATA_EXFILTRATION]: The tool atlas_upload_media allows the agent to read local files from an absolute path and upload them to Atlas Cloud's infrastructure. While intended for media generation, this capability grants the agent access to the local file system.
[EXTERNAL_DOWNLOADS]: The skill dynamically fetches model lists and OpenAPI schemas from api.atlascloud.ai at runtime to ensure compatibility with the current API version.

atlas-cloud