The Agent Skills Directory

[SAFE]: The skill performs routing logic by inspecting file extensions and directory paths within the project workspace. It lacks dangerous primitives such as shell execution, file modification, or network access.
[DATA_EXPOSURE]: The skill reads project-specific files including qx_method_candidates.md, model-code-analyzer.md, and source code files. These accesses are restricted to the project context and do not target sensitive system locations.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from generated code and documentation files to determine routing. The risk is considered safe as the skill lacks exploitable capabilities like command execution or file-writing primitives.

code-reviewer