The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from external files which may contain malicious instructions intended to hijack the agent's behavior.
Ingestion points: The skill instructions in SKILL.md specify reading content from a file path provided via $ARGUMENTS.
Boundary markers: No delimiters or protective instructions are used to separate the draft content from the agent's own task instructions.
Capability inventory: The skill has the capability to read local files and launch subordinate agents (SKILL.md).
Sanitization: There is no validation or filtering performed on the file content before it is processed.
[DATA_EXFILTRATION]: The skill allows for reading files from arbitrary paths supplied by the user. Without restricted scoping or directory validation, this capability could be used to read and expose sensitive system information, configuration files, or credentials to the agent's context.

quality