The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from documentation files (e.g., docs/*.mdx) for analysis. Maliciously crafted text in these files could attempt to influence the agent's review logic or suggested edits. This represents a standard attack surface for document analysis tools.
Ingestion points: Raw Markdown/MDX content is read from user-provided files within the docs/ directory.
Boundary markers: The skill does not implement explicit boundary markers or security delimiters when passing ingested content to sub-agent workers.
Capability inventory: The skill has the capability to write edited text back to the documentation files and update local log files in the logs/ directory.
Sanitization: There is no specific evidence of input sanitization or filtering to prevent embedded instructions in the documentation from being executed by the analysis agents.

review-doc