The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source code during its auditing and simplification phases, creating a surface where malicious instructions embedded in the code could influence the agent's behavior. * Ingestion points: Project source code files are accessed through search and read operations in SKILL.md. * Boundary markers: There are no specific instructions or delimiters provided to prevent the agent from obeying commands found within the analyzed files. * Capability inventory: The agent is authorized to use grep_search and replace_file_content (SKILL.md). * Sanitization: No logic is defined to sanitize or filter out natural language instructions found in the codebase.
[COMMAND_EXECUTION]: The skill is explicitly designed to perform 'Active Refactoring' in its SIMPLIFY mode, which involves direct modification of the project's source code using the replace_file_content tool. While this is the primary feature of the skill, it represents a high-impact capability that modifies the user's filesystem.

code-auditing