The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides instructions that direct the agent to actively evaluate and potentially remove defensive coding practices. It explicitly tells the agent to perform 'mental ablation' on guards, validation, and exception handlers, and to 'remove what adds no concrete value.' This behavior could lead the agent to strip away security controls in the code it generates based on its own subjective assessment of risk.
[PROMPT_INJECTION]: The instructions discourage the use of 'defensive systems built around hypothetical same-machine tampering' and 'repeated manifest consistency checks.' This may influence the agent to ignore or disable security features designed to protect against local threats or ensure file integrity, particularly if the agent misinterprets the realistic risks of the operating environment.

proportionate-engineering