The Agent Skills Directory

[SAFE]: No malicious patterns, data exfiltration, or credential exposure were detected. The skill consists exclusively of markdown documentation, architectural guidelines, and templates for capturing design rationale.\n- [NO_CODE]: The skill does not contain any executable scripts, binaries, or runtime logic. It specifies 'No runtime dependency' and relies solely on the agent's ability to read and write text files according to the provided instructions.\n- [PROMPT_INJECTION]: Analysis of 'references/fitness-functions.md' identified a prompt template for 'AI-Assisted Fitness Functions' that uses directive markers like 'IMPORTANT:' and model feature commands. These are presented as a legitimate methodology for enhancing automated technical reviews and do not attempt to bypass agent safety filters or override system instructions.

adr-authoring