The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains phrases like 'ignore previous instructions' in references/opportunities.md and references/improve.md. These are defensive instructions directed at the agent to ensure it treats analyzed source code as inert data, preventing it from being manipulated by malicious comments or code in the target repository. This is a security best practice for agents that process external data.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external codebases and diffs, creating an attack surface for indirect prompt injection. This is explicitly addressed and mitigated by 'Hard Rule 4' in the workflow files, which instructs the agent to flag and ignore any instruction-steering patterns found in analyzed files, maintaining a strict boundary between data and instructions.
[REMOTE_CODE_EXECUTION]: While the skill discusses various animation libraries and implementation strategies, it does not contain any instructions or scripts that perform automated downloads or execution of remote code. All external references are to well-known design and documentation resources.

kernel-animate