The Agent Skills Directory

[COMMAND_EXECUTION]: The file-organizer component executes powerful shell commands including find, du, mv, and mkdir to restructure files and directories. It is explicitly designed to operate on the user's home directory and other high-sensitivity filesystem paths.
[COMMAND_EXECUTION]: The skill-generator framework implements a script execution engine that constructs shell commands for Bash by concatenating runtime paths and arguments. This assembly method is vulnerable to shell command injection as it relies on simple double-quoting without robust sanitization of metacharacters in the arguments.
[COMMAND_EXECUTION]: The skill facilitates meta-programming by allowing the agent to generate and execute new Python and Bash scripts at runtime. This enables the agent to create and load entirely new behaviors and toolsets into the environment.
[PROMPT_INJECTION]: There is a significant surface for indirect prompt injection. The agent reads untrusted data from filenames, file content, and metadata to drive organizational logic and code generation, which could allow malicious data to influence the agent's task execution.
[DATA_EXFILTRATION]: Through the file-organizer, the agent can traverse the local filesystem to calculate cryptographic hashes (MD5) and summarize directory usage. This provides a mechanism for exposing sensitive metadata and file content patterns to the agent's context.

operations-specialist