The Agent Skills Directory

[SAFE]: The skill consists entirely of natural language instructions for an interviewing process. It does not contain any executable scripts, binaries, or remote dependencies.
[PROMPT_INJECTION]: The directives to "Interview the user relentlessly" and "Treat the user's correction as the new truth immediately" are functional requirements for the skill's intended purpose of stress-testing plans. These do not constitute malicious attempts to bypass safety filters or extract system prompts.
[DATA_EXPOSURE]: The instruction to "explore files, docs, or code" to resolve questions leverages the agent's existing workspace access in a standard way and does not target sensitive configuration files or credentials.

grilling