The Agent Skills Directory

[DATA_EXFILTRATION]: The skill allows the agent to read any file on the filesystem and transmit its contents to a remote server. Because there are no restrictions on which files or directories can be accessed, a malicious actor or an accidental prompt could lead to the exfiltration of sensitive data such as SSH keys, environment secrets, or configuration files.\n- [COMMAND_EXECUTION]: The skill uses curl to perform network operations. It dynamically incorporates the file path ($file) and environment variables ($SHAREFLARE_URL, $SHAREFLARE_TOKEN) into the command, enabling the transfer of local data to an external network.\n- [PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by processing external data and possessing network-enabled capabilities.\n
Ingestion points: Files requested by the user or agent, accessed via the file system in SKILL.md.\n
Boundary markers: None. The instructions do not define safe versus unsafe file locations or content.\n
Capability inventory: Shell command execution for file uploading and metadata retrieval in SKILL.md.\n
Sanitization: Absent. Files are uploaded exactly as they exist on disk.

shareflare