The Agent Skills Directory

[COMMAND_EXECUTION]: The instruction to execute recall search "<query>" with user-supplied text allows for shell command injection. Malicious input containing characters like ;, &, or $(...) can lead to arbitrary code execution on the host system.
[PROMPT_INJECTION]: The skill ingests historical memory data which may contain malicious instructions, creating an indirect prompt injection vulnerability.
Ingestion points: Search results from the recall command.
Boundary markers: None.
Capability inventory: The agent executes the recall CLI tool.
Sanitization: No validation or escaping of input or output is specified.

recall-search