The Agent Skills Directory

[PROMPT_INJECTION]: The skill's description and body use directive language like 'Stop.' and 'Wait, I don't understand...' to redirect agent behavior, which mimics prompt injection techniques.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the environment and previous interactions.
Ingestion points: The instructions in SKILL.md reference the 'last message' and an external CONTEXT.md file.
Boundary markers: No delimiters or isolation instructions are provided to separate the ingested data from the skill's operational instructions.
Capability inventory: No dangerous tools, subprocess calls, or network operations were found in the provided files.
Sanitization: The skill does not perform any validation or sanitization of the external content before processing it.

meaning