The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from potentially untrusted external sources, such as model outputs, web scrapes, and OCR data, which could contain malicious instructions.
Ingestion points: Data enters the context via router call chat, router call scrape, router call transcribe, and router call ocr commands as described in references/olares-router-calling.md and references/olares-router-local-multimodal.md.
Boundary markers: No specific delimiters or safety warnings for the agent to ignore instructions in processed data are mentioned in the skill files.
Capability inventory: The agent has the ability to execute various olares-cli commands, perform network operations, and access local files.
Sanitization: The skill does not define methods for sanitizing or validating ingested data before it is processed by the agent.
[COMMAND_EXECUTION]: The skill uses the olares-cli binary to perform system-level operations, including model management and configuration. These actions are within the expected scope for an administrative gateway tool.

olares-router