The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill has an inherent surface for indirect prompt injection as it reads and processes user-provided code files. An attacker who can influence the project's source code could potentially embed instructions aimed at the agent. However, the skill's capabilities are limited to standard file reading and writing within the project scope. Ingestion points occur in Step A3 (Analyse mode) and Step D1 (Decompose mode) where the agent reads local files. There are no explicit boundary markers or sanitization steps mentioned in the skill instructions. The agent's primary capabilities include generating and refactoring component files based on the analyzed code.
[SAFE]: No malicious patterns such as prompt injection, obfuscation, or privilege escalation were detected. The skill's behavior is consistent with its stated purpose of improving React code quality. All references are local markdown files providing legitimate architectural guidance, and no remote dependencies or network calls are utilized.

react-composition-rules