Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed QVeris API integration that sends tool-discovery queries and tool-call parameters to qveris.ai using the configured API key.
Install this only if you are comfortable sending relevant task details and tool parameters to QVeris under your QVERIS_API_KEY. Avoid using it for secrets or sensitive personal, medical, financial, or proprietary data unless that use is acceptable under your QVeris account and data handling expectations.
The skill enables outbound network access and describes multiple execution paths, but it does not declare a restrictive tool scope such as allowed-tools or permissions. In an auto-invoked skill, this broadens what the agent may attempt at runtime and reduces policy enforcement, increasing the chance of unintended external calls or tool misuse.
The description states that discovery queries 'must be English API capability descriptions,' which imposes a specific language requirement. This is a natural-language policy concern because it removes user language choice rather than framing English as an implementation detail with an opt-in or fallback.
The markdown says user requests in any language 'must be converted to English' for discovery queries. Even if intended for backend compatibility, this is written as a mandatory language constraint and does not offer user choice or clearly limit the requirement to internal processing only.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
| Wrong date format | `"date": "01/15/2026"` | `"date": "2026-01-15"` (ISO 8601) |
| Missing required param | Omitting `symbol` for a stock API | Always check required list |
| Natural language or wrong format as param | `"query": "what is AAPL price"` or `"symbol": "Apple"` | Extract structured values: `"symbol": "AAPL"` |
| Constructing API URLs manually | Directly calling `https://api.qveris.com/...` | Use the API reference above or the script |
| Giving up after one failure | "I don't have real-time data" / abandoning after error | Discover first; follow Error Recovery on failure |
| Not trying http_request when exec fails | Abandoning when node/exec is unavailable | Use http_request tool (Tier 2) — it works without exec |
This code sends user-provided query text, tool identifiers, discovery IDs, and execution parameters to a remote API via fetch, but it contains no confirmation prompt, logging, comments, or docstrings that disclose this external transmission. Because the file is code and the operation involves network transfer of potentially user or system data, it matches the missing-warning criterion.
No suspicious patterns detected.