The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill interacts with untrusted external data, which presents a surface for indirect prompt injection attacks where malicious content in URLs or torrent files could attempt to subvert agent instructions.
Ingestion points: Processes arbitrary user-supplied URLs through create, inspect, and file exists commands; ingests local BitTorrent files via torrent inspect and create --torrent as described in references/olares-knowledge-download-torrent.md.
Boundary markers: The skill explicitly instructs the agent in SKILL.md to seek user confirmation before performing actions such as creating, cancelling, or removing tasks, and before executing file removals or settings changes.
Capability inventory: The skill possesses capabilities including initiating network downloads, removing files within the Olares persistent volume (drive/Home/...), and managing session credentials via cookie and token imports.
Sanitization: Instructions specify quoting parameters to handle special characters in URLs and note that the CLI performs local validation of status enums and file indices.
[COMMAND_EXECUTION]: The skill relies on executing the olares-cli binary to perform all download management tasks. The instructions define specific subcommand structures and parameters for interacting with the Olares platform services.

olares-knowledge