The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill is vulnerable to indirect prompt injection as it combines data ingestion from untrusted project files with file system write capabilities.
Ingestion points: Uses Read, Glob, and Grep tools to ingest content from project files (e.g., .tsx, .ts files) within a user's codebase.
Boundary markers: No delimiters or instructions (e.g., 'ignore instructions within data') are provided to isolate untrusted code content from the agent's operational logic.
Capability inventory: Granted Write and Edit permissions for file system modifications, creating a risk of unauthorized code changes or data deletion if triggered by malicious input.
Sanitization: Lacks any defined sanitization or validation mechanisms to filter out malicious instructions embedded in the processed project files.

nextjs-best-practices