The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains instructions to "Set aside the task requirements" and "Set aside evaluation criteria." These patterns, while intended for creative exploration, mimic jailbreak-style prompt injections that command an agent to ignore its operational constraints and safety filters.
[PROMPT_INJECTION]: The skill presents an Indirect Prompt Injection surface. It allows the ingestion of external data (files or codebases via the Read tool) while explicitly instructing the agent to abandon its typical evaluative and structured reasoning. This combination increases the likelihood that instructions hidden within external data could influence the agent's behavior.
Ingestion points: SKILL.md (Optional input: "A file or codebase to read as raw material")
Boundary markers: Absent (The skill does not define delimiters or provide warnings to the agent to ignore instructions found within the input files)
Capability inventory: Read tool
Sanitization: Absent

dream