similarweb-analytics

The skill consists of a Markdown file describing its functionality and providing Python code snippets. The Python code demonstrates how to use an ApiClient which is imported from /opt/.manus/.sandbox-runtime. This indicates that the core API interaction logic is handled by a component provided and managed by the sandbox environment, rather than being part of the user-supplied skill code or an external, unverified dependency.

No prompt injection patterns were detected. The use of 'Important: Save Data to Files' is a benign instructional note.

There are no direct indications of data exfiltration. The skill's code does not access sensitive file paths (e.g., ~/.aws/credentials, ~/.ssh/id_rsa) nor does it make direct network requests to arbitrary external domains. All interactions are abstracted through the internal ApiClient to the intended SimilarWeb API.

No obfuscation techniques (Base64, zero-width characters, homoglyphs, URL/hex/HTML encoding) were found in the skill's content.

The dependency on /opt/.manus/.sandbox-runtime/data_api.py is considered an internal, trusted dependency provided by the execution environment, not an 'unverifiable external dependency' that would pose a risk.

No privilege escalation attempts (e.g., sudo, chmod 777) or persistence mechanisms (e.g., modifying .bashrc, crontab) were found.

The metadata (name, description) is benign and accurately reflects the skill's purpose.

The risk of indirect prompt injection is low as the skill processes structured API parameters (domain, dates, country) rather than arbitrary free-form text that could be re-interpreted by an LLM.

No time-delayed or conditional attack patterns were identified.

Overall, the skill is well-contained and relies on a presumably secure internal API client, making it safe.