The Agent Skills Directory

[SAFE]: The skill provides instructions for integrating official Univer Pro features using verified vendor packages from the developer 'dream-num', such as @univerjs-pro/license and @univerjs-pro/collaboration.
[SAFE]: Security best practices are enforced regarding credential management; the instructions explicitly warn against committing client licenses to source control or embedding server licenses in frontend bundles.
[SAFE]: Legitimate product functionality, including file exchange via Universer endpoints and browser worker initialization for performance optimization, is documented using standard web development patterns.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external data which could potentially harbor malicious instructions.
Ingestion points: External documents are imported into the agent context through methods such as importSheetToUnitIdAsync (SKILL.md) and importDocToSnapshotAsync (references/exchange-guide.md).
Boundary markers: The provided documentation does not include specific instructions to use delimiters or ignore-previous-instruction markers when the agent processes these files.
Capability inventory: The integrated library features capabilities for file downloads (downloadFile), network requests to collaboration/exchange endpoints, and runtime plugin registration.
Sanitization: There are no documented steps for the agent to sanitize or validate the content of imported documents for indirect prompt injection vectors.

univer-pro-integrate