The Agent Skills Directory

[SAFE]: The skill is composed of instructions, prompts, and evaluation cases designed for technical design review. It does not include any executable scripts, binary files, or commands that interact with the host system's sensitive files or network.
[SAFE]: The instructions in prompts/error-handling-design-review.md and SKILL.md explicitly forbid the agent from performing dangerous actions such as running fault injection, reviewing real incidents in production, or making autonomous risk acceptance decisions on behalf of a human user.
[SAFE]: No patterns of prompt injection, data exfiltration, or obfuscation were detected. All external references point to local configuration files used for testing and validation within the skill's own directory structure.

error-handling-design-review