The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted code and documentation from the local repository, creating a potential attack surface for indirect prompt injection. Malicious instructions embedded within target files could influence the agent's decision-making process during the automated refactoring and review cycles.
Ingestion points: File content is read via the Read tool in SKILL.md and references/target-detection.md for target classification and complexity scoring.
Boundary markers: None defined. The instructions do not utilize specific delimiters or isolation techniques to separate data being processed from the agent's internal logic.
Capability inventory: The skill possesses high-privilege capabilities including Edit, Write, and Bash for filesystem modification and analysis.
Sanitization: Path validation is correctly implemented in references/target-detection.md to prevent unauthorized file access, but there is no sanitization of file content for potential embedded instructions.

refactor