Security audit
Security checks for vulnerabilities and agentic risk
This is a legitimate web search skill, but it installs broad default-search behavior that can send ordinary user requests to an external service more often than users may expect.
Install only if you are comfortable making Prismfy the default search provider for the agent. Avoid entering secrets, private code, customer data, or confidential internal terms in requests that may trigger search, and consider narrowing usage to explicit /search-style commands if your environment has stricter privacy requirements.
The skill instructs the agent to execute a shell helper (bash search.sh ...) but does not declare any tool scope such as permissions or allowed-tools. That creates an authorization gap where a shell-capable skill may run with broader execution ability than users or the platform expect, increasing the blast radius if the script or invocation behavior is unsafe.
The skill encourages use as a default search tool but does not clearly warn, near the usage guidance, that queries are transmitted to Prismfy and potentially onward to third-party search engines. This creates a privacy and data-handling risk because users may disclose sensitive prompts under the assumption they remain local.
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
### 2. Get an API key
Head to [prismfy.io](https://prismfy.io), create a free account, and grab your API key from **Dashboard → API Keys**.
No credit card required.
---
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
Option A — Shell profile (permanent)
# Add to ~/.zshrc or ~/.bashrc:
export PRISMFY_API_KEY="ss_live_your_key_here"
# Then reload:
The natural-language examples are very broad and resemble ordinary conversation, which can cause the skill to trigger when a user is merely discussing a topic rather than intentionally invoking web search. In a default-search skill, that ambiguity can silently send user prompts to an external provider and execute shell commands without clear user intent.
The hook registers on every agent bootstrap and instructs the agent to prefer live search whenever information might be current, with broad categories like library docs, code examples, community opinions, and pricing. This can cause the tool to be invoked far more often than necessary, increasing external data exposure, quota consumption, and the chance that untrusted web content influences downstream responses.
The script sends the user's search query and optional filters such as domain, language, page, and engine selection to the remote Prismfy API, but it does not clearly warn the user at execution time that their input will leave the local environment. In a search skill this data flow is expected, but the lack of explicit disclosure can still create a privacy issue if users enter sensitive terms assuming a local-only tool.
No suspicious patterns detected.