The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents a pattern where user input is processed via the Perplexity Sonar API, which performs live web searches. This creates an inherent surface for indirect prompt injection from untrusted web content.
Ingestion points: The rawQuery parameter is ingested in the safeSearch function in SKILL.md.
Boundary markers: The skill does not demonstrate the use of delimiters or specific system instructions to isolate the search query from potential instructions in search results.
Capability inventory: The skill uses the perplexity.chat.completions.create tool.
Sanitization: The skill provides a robust sanitizeQuery function to redact PII (emails, phone numbers, SSNs, and credit cards) and API keys, which effectively mitigates the primary data exposure risks mentioned in the guidance.
[EXTERNAL_DOWNLOADS]: The skill includes a curl example targeting the official Perplexity API domain (api.perplexity.ai) for credential verification, which is a well-known and expected service for this context.

perplexity-security-basics