grammarly-data-handling
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a file processing pipeline that reads content from local files, creating a surface for indirect prompt injection from untrusted data.
- Ingestion points: The
scoreFilefunction inSKILL.mdusesfs.readFileSyncto ingest content from local file paths. - Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are implemented in the provided code snippets when handling text chunks.
- Capability inventory: The skill utilizes the Node.js
fsmodule to access the filesystem and provides a pattern for passing data to an external API client. - Sanitization: The input text is not sanitized or escaped before being processed by the scoring logic.
- [DATA_EXPOSURE]: The skill provides patterns for reading arbitrary files from the filesystem using
fs.readFileSync. While appropriate for its stated purpose of document processing, this functionality allows access to potentially sensitive local data if the environment does not restrict accessible paths.
Audit Metadata