The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection due to its core functionality of reading and analyzing untrusted external content.
Ingestion points: The skill utilizes Read, Grep, and Glob tools to ingest and process external source code and pull request data (SKILL.md).
Boundary markers: There are no specific instructions or delimiters defined to prevent the agent from following instructions that might be embedded within comments, documentation, or string literals in the code being reviewed.
Capability inventory: The skill is limited to read-only file operations (Read, Grep, Glob) and does not possess capabilities for file writing, network communication, or arbitrary command execution.
Sanitization: No sanitization or filtering mechanisms are implemented to distinguish between code-under-review and potential instructions embedded within that code.

code-review