The Agent Skills Directory

[SAFE]: Analysis of the skill instructions, architectural guidelines, and templates revealed no evidence of malicious intent. The skill promotes industry-standard security and development practices, such as dependency inversion and role-based interface segregation.
[INDIRECT_PROMPT_INJECTION]: The skill requires the agent to read existing codebase files to ensure DRY (Don't Repeat Yourself) compliance and architectural consistency. While reading existing source code represents a theoretical surface for indirect prompt injection, the risk is categorized as safe due to the skill's enforcement of strict architectural boundaries and small file sizes.
Ingestion points: Reads existing project source files via grepping logic and specialized architectural exploration agents.
Boundary markers: Enforces a strict modular directory structure (modules/[feature]/) and aggressive line limits (max 100 lines per file) that restrict the complexity of ingested content.
Capability inventory: File system read and write access for local Java source files, directory layout management, and configuration wiring.
Sanitization: Relies on strict adherence to predefined, restricted code templates and modular isolation to prevent malicious instructions from propagating through the codebase.

solid-java