The Agent Skills Directory

[COMMAND_EXECUTION]: The skill suggests using shell commands like grep or ripgrep (rg) to identify typography errors. These patterns are used for local searching and do not involve executing external code or scripts.
[PROMPT_INJECTION]: The skill processes untrusted user-provided text (ingestion point) without explicit boundary markers or sanitization. However, the instructions mandate narrow, mechanical corrections based on a static reference file, which minimizes the ability of injected instructions to influence the agent's behavior.
[SAFE]: The skill reads a local configuration file (brand-brief.md) to adapt its corrections to specific brand styles. This is a standard practice for project-specific context and poses no security threat.

typografie-de