The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted data from code comments.
Ingestion points: The skill reads external code content using the Read and Grep tools.
Boundary markers: There are no specific delimiters or instructions provided to the agent to distinguish between the analyzed text and its own instructions, nor any directive to ignore embedded commands.
Capability inventory: The skill has access to Bash and TodoWrite tools, providing a path for command execution or file system modification if an injection is successful.
Sanitization: No sanitization, escaping, or validation logic is defined for the input content.
[COMMAND_EXECUTION]: The skill explicitly allows the use of the Bash tool. While intended for code analysis tasks, this capability provides a powerful execution environment that could be abused by malicious content found within the code comments being analyzed.

comment-analyzer