The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the local file system (filenames, file metadata, and directory structures) which could potentially contain malicious instructions intended to override agent behavior.
Ingestion points: Commands like ls, find, du, and file read names and headers of files within user-specified directories (e.g., in SKILL.md).
Boundary markers: The instructions do not include specific delimiters or instructions to treat file system data as untrusted text, which could lead the agent to follow instructions embedded in filenames.
Capability inventory: The agent is granted the ability to create directories (mkdir), move files (mv), and delete duplicates (rm context) via shell commands.
Sanitization: There is no mention of sanitizing or escaping filenames or command outputs before they are processed by the agent's logic.

file-organizer