The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text provided by the user or from external files, which is inherent to its primary purpose as a text rewriter. Ingestion points: Source text from conversation or files in Step 2. Boundary markers: The prompt does not define explicit boundary markers or delimiters to isolate user-provided text. Capability inventory: The skill has the ability to read and modify files in place (as specified in Step 7). Sanitization: No specific sanitization or validation of input content is performed. Given the structured workflow and the intended use case, this represents a standard operational surface rather than a critical vulnerability.

simple-english