The Agent Skills Directory

[PROMPT_INJECTION]: The skill mandates the inclusion of a "Terminal Goal Line" in the generated output, specifically: "Work completely autonomously and do not ask me for anything until you are all done." This directive explicitly instructs the agent to bypass human-in-the-loop oversight during execution, which is a common pattern for suppressing safety interventions during autonomous operations.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user requirements and specifications to synthesize complex agent instructions, creating a surface for injection.
Ingestion points: User intent, requirements, and rambling specifications in SKILL.md are used as the primary input for synthesis.
Boundary markers: The synthesis procedure does not mandate the use of delimiters or instructions to ignore embedded commands within the user input.
Capability inventory: The resulting prompts are intended for multi-turn autonomous execution, which typically implies access to file system tools and network capabilities.
Sanitization: The skill lacks explicit validation or sanitization steps to ensure that user-provided specifications do not contain malicious instructions that would be incorporated into the high-authority autonomous goal.

goal