The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands within the Phase 2 (Mastery): Core Loop section. In the 해설하기 (REVEAL) step, the instructions state that if 'Code Execution Mode' is enabled, the agent must run the Swift code snippets using the swift compiler via a bash tool to display output or compiler errors to the user.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it processes user-supplied topics to generate content that is subsequently executed.
Ingestion points: User input provided during the Topic Selection phase in SKILL.md is used to define the learning subject.
Boundary markers: There are no delimiters or specific warnings to the model to ignore instructions embedded within the user-provided topic.
Capability inventory: The agent possesses the capability to execute shell commands via the bash tool to run the swift compiler.
Sanitization: The instructions do not include specific sanitization or validation logic to ensure that a user-provided topic cannot influence the agent to generate and execute harmful code.

swift-study