Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed local Perplexica search helper with no evidence of hidden persistence, credential theft, destructive behavior, or remote payload execution.
Install this only if you run and trust a local Perplexica instance. Treat searches as potentially leaving your machine through Perplexica's configured search and LLM providers, and use explicit Perplexica wording for sensitive queries to avoid unintended activation.
The skill clearly performs network actions against a local Perplexica HTTP API, but the manifest does not declare an explicit tool scope such as permissions or allowed-tools. Even though the documentation describes local-only safeguards, the missing scope weakens enforcement at the platform level and can cause the runner to grant broader-than-necessary network capability or make review harder.
The trigger phrases are broad enough that an agent may invoke this skill for generic requests like 'deep search' or 'research with sources' even when another tool would be more appropriate. This can create unintended networked behavior, send user queries into Perplexica without sufficient user intent confirmation, and increase the chance of over-activation in sensitive contexts.
No suspicious patterns detected.
