The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill uses the $ARGUMENTS variable to ingest user input directly into its core instructions without any safety delimiters or sanitization. This establishes a vulnerability surface where malicious instructions embedded in the data could influence the agent's behavior.
Ingestion points: User-supplied content is passed into the skill via $ARGUMENTS in the SKILL.md file.
Boundary markers: No delimiters, such as triple backticks or specific XML-style tags, are used to separate user data from the skill's operational instructions.
Capability inventory: The skill instructions explicitly authorize and direct the agent to use the Bash tool and Task tool for background execution.
Sanitization: There is no logic provided to escape, validate, or filter the content of $ARGUMENTS before it is passed to powerful system-level tools.
[COMMAND_EXECUTION]: The primary purpose of this skill is to facilitate the execution of shell commands via the Bash tool. By providing a streamlined way to run these commands in the background without direct user oversight of the execution process, it increases the risk of the agent performing unintended or malicious system operations.

background