The Agent Skills Directory

[SAFE]: The skill implements robust pre-flight checks, requiring the agent to verify authentication via auth:check and confirm the existence of collections and fields before performing any queries.
[SAFE]: Data access is confined to the authenticated NocoBase MCP environment. The skill instructions prioritize the main data source and require explicit identification of data sources for all resource calls, preventing unauthorized or accidental data access across environments.
[SAFE]: The skill provides detailed guidance on validating query contracts (e.g., using aggregation instead of aggregate) and recommends cross-checking aggregate results with manual counts from record lists to ensure data integrity and prevent misinterpretation of results.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze potentially untrusted business data from NocoBase collections. While this presents a surface for indirect prompt injection, the skill's focus is on generating summaries and breakdowns. No high-risk downstream capabilities, such as arbitrary command execution, file system writes, or external network exfiltration, were identified that would allow a malicious payload in the data to compromise the agent or environment.

nocobase-data-analysis