The Agent Skills Directory

[SAFE]: The skill is strictly scoped to Java language conventions (concurrency, exceptions, NIO, etc.) and explicitly delegates security and architecture reviews to other specialized skills, adhering to the principle of least privilege.
[SAFE]: No remote code execution, external network requests, or sensitive file access patterns were detected. The skill uses a structured findings-list output schema and does not involve subprocess calls or shell execution.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface as it ingests untrusted Java source code through the code-scope input.
Ingestion points: Java source files or directories provided via the input_schema in SKILL.md.
Boundary markers: None explicitly defined to separate user code from agent instructions.
Capability inventory: The skill is limited to generating textual findings (location, category, severity, suggestion) and does not have write access to the filesystem, network access, or dynamic execution capabilities.
Sanitization: Not present. While the lack of boundaries allows for potential prompt injection via code comments, the lack of powerful tools or capabilities associated with this skill renders the risk negligible.

review-java