The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill interfaces with the perplexity command-line tool. It implements the principle of least privilege by restricting the agent's environment to Bash(perplexity:*) commands only.
[DATA_EXFILTRATION] (SAFE): While the tool transmits data to Perplexity AI, this is consistent with its stated purpose of performing AI searches. No unauthorized data exfiltration to untrusted domains was identified.
[PROMPT_INJECTION] (LOW): The skill is susceptible to indirect prompt injection because it can process instructions from untrusted external data sources like local files and standard input.
Ingestion points: Usage of -f flag for file reading and shell piping (e.g., cat questions.txt | while read q).
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined for processed content.
Capability inventory: The skill possesses the Bash tool for command execution and Read for file access.
Sanitization: There is no evidence of input sanitization or validation before content is passed to the CLI tool.

perplexity-cli