The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to interact with the host system by executing commands via the sau CLI utility.
Evidence: The skill body and example scripts (scripts/examples/facebook_commands.sh, scripts/examples/facebook_commands.ps1) contain shell commands for logging in and uploading content.
Evidence: scripts/examples/facebook_cli_template.py uses subprocess.run() to execute these commands programmatically.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, such as video titles, descriptions, and post URLs, which are then passed as arguments to shell commands. This creates a surface where malicious instructions embedded in the processed content could attempt to influence the agent's behavior.
Ingestion points: Input parameters for the sau command including --title, --desc, --tags, and --post-url (found in SKILL.md and script templates).
Boundary markers: None identified; arguments are passed directly to the CLI tool.
Capability inventory: The skill has the capability to execute shell commands and write data to the local filesystem (e.g., saving session cookies to cookies/facebook_creator.json).
Sanitization: No explicit sanitization or validation of the input strings is performed in the provided example scripts.

facebook-upload