The Agent Skills Directory

[COMMAND_EXECUTION]: The virlo.sh script modifies shell startup files including ~/.bashrc, ~/.zshrc, and ~/.bash_profile to persist the VIRLO_API_KEY environment variable. While this is presented as a setup convenience, modifying shell profiles is a persistence mechanism that changes the user's shell environment for future sessions.
[COMMAND_EXECUTION]: The skill performs dynamic code execution by piping logic into python3 using both -c command-line arguments and shell heredocs. This is used for complex JSON manipulation, file editing, and scoring logic within the virlo.sh script.
[PROMPT_INJECTION]: The skill fetches content such as video descriptions and creator metadata from TikTok, Instagram, and YouTube via the virlo.ai API. This external, untrusted data is returned to the agent and is intended for downstream analysis, presenting a surface for indirect prompt injection attacks where malicious instructions could be embedded in the social media content.
Ingestion points: API responses fetched via the videos, outliers, and ads commands in scripts/virlo.sh.
Boundary markers: None observed; content is processed and returned without explicit delimiters or warnings to ignore embedded instructions.
Capability inventory: curl, jq, python3, sed, tail, mkdir, printf, uname, date, read (found in scripts/virlo.sh).
Sanitization: None observed; the script parses and filters the data but does not sanitize the text content for potential prompt instructions.

virlo