The Agent Skills Directory

[SAFE]: The skill implements a straightforward Python client to interface with Perplexity Sonar search endpoints. The code uses standard libraries and follows expected patterns for API-based tools.
[DATA_EXFILTRATION]: The client communicates with api.aisa.one. This domain is the official infrastructure for the skill author (aisa-team), and all network traffic is directed toward these legitimate service endpoints.
[CREDENTIALS_UNSAFE]: The skill correctly manages authentication by requiring the AISA_API_KEY to be provided via an environment variable, avoiding hardcoded secrets.
[INDIRECT_PROMPT_INJECTION]: As a search and research tool, the skill naturally ingests user-provided strings (query and system instructions) to be processed by an external LLM (Perplexity).
Ingestion points: User input provided via the --query and --system arguments in scripts/perplexity_search_client.py.
Boundary markers: Arguments are encapsulated within a structured JSON payload sent to the API.
Capability inventory: The skill is limited to performing network POST requests to the vendor's API and returning the resulting text.
Sanitization: Standard JSON serialization is used for data transmission; downstream content is handled by the Perplexity engine's own safety guardrails.

perplexity-search