The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/check-versions.sh script fetches metadata from PyPI and NPM registries to check for version drift. These network operations target official package registries and are used for maintenance purposes.
[COMMAND_EXECUTION]: The skill leverages MCP tools to perform backend operations, including executing raw SQL and registering serverless functions. These are high-privilege capabilities necessary for the skill's role as a resource provisioning control plane.
[PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection where untrusted data from datatables or storage buckets could influence the agent's behavior. Ingestion points: Data is retrieved via datatable_data and manage_storage (SKILL.md). Boundary markers: No explicit instruction-level delimiters or 'ignore embedded instructions' warnings are provided in the prompt logic. Capability inventory: High-privilege actions like execute_raw_sql, manage_policies, manage_function, and create_user are available. Sanitization: Security validation is performed at the tool level (e.g., SQLAnalyzer for raw SQL), but no specific NLP-level sanitization for ingested data is documented.

taruvi-backend-provisioning