The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests untrusted data from the user's project directory during its analysis phase.\n
Ingestion points: The skill uses Glob and Grep tools to scan project files including **/privacy*.md, **/terms*.md, **/eula*.md, and various source code files to identify SDK usage and permission descriptions in Info.plist.\n
Boundary markers: There are no explicit delimiters or specific instructions for the agent to ignore or isolate instructions that may be embedded within the read files.\n
Capability inventory: The skill has the ability to read, write, and edit local files using Read, Write, Edit, Glob, and Grep tools, allowing it to modify the project based on the analyzed data.\n
Sanitization: Extracted content is not sanitized or validated before being incorporated into the final legal documents generated by the skill.

privacy-policy