The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data in the form of source code and git diffs, which represents a potential surface for indirect prompt injection.
Ingestion points: As described in SKILL.md, the agent accepts input through 'Git Diff' mode (reading changed files), 'File Review' mode (reading specified files), and 'Snippet Review' mode (processing code directly provided by the user).
Boundary markers: The skill documentation does not define specific boundary markers (such as XML tags or unique delimiters) to separate its instructions from the untrusted code being reviewed, which could theoretically allow instructions inside a code snippet to influence the agent's behavior.
Capability inventory: The skill is primarily focused on text analysis and reporting. It does not contain scripts or instructions that perform sensitive operations like network exfiltration, arbitrary shell execution, or persistent file system modifications.
Sanitization: No explicit sanitization or filtering logic is specified for the input data before it is processed by the agent.
[SAFE]: The skill uses clear, well-documented guidelines for code quality and security (found in references/) and follows best practices for providing structured feedback without requesting elevated privileges or external dependencies.

code-review