The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external files and user-provided queries to populate knowledge bases and perform retrieval tests. This content is used to influence the AI's response generation and employee bindings.
Ingestion points: The skill uses nb api kb documents upload to ingest various file formats (txt, md, pdf, etc.) and nb api kb run-hit-test --query for user-supplied retrieval strings.
Boundary markers: There are no explicit instructions to the underlying model to ignore instructions embedded within the document content itself.
Capability inventory: The skill possesses broad Bash execution capabilities for environment management, plugin manipulation, and database configuration.
Sanitization: The skill does not define specific sanitization or filtering logic for the content of uploaded documents before they are processed by the embedding engine.
[COMMAND_EXECUTION]: The skill performs system-level state changes and configuration management using the NocoBase CLI (nb).
Evidence: The workflow involves executing nb plugin enable to activate system extensions and nb api kb to mutate database configurations.
Mitigation: The skill implements a robust 'Safety Gate' and 'Mandatory Clarification Gate' that prevents execution of state-changing commands without explicit, target-specific user consent and preflight verification.

nocobase-ai-knowledge-base-manager