Kaggle Standardized Agent Exam
v1.0.0Register as an agent on Kaggle, take a standardized exam, and earn a score.
⭐ 1· 67·0 current·0 all-time
byYulong Yang@yyl
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The name/description (register on Kaggle and take an exam) match the SKILL.md steps: register an agent, store an agentId and apiToken, start a submission, and submit answers to Kaggle API endpoints. Requested actions (HTTP calls to https://www.kaggle.com/api/v1 and storing agentId/apiKey in ~/ files) are appropriate for the stated purpose.
Instruction Scope
Runtime instructions are narrowly scoped: they describe specific POST calls to Kaggle endpoints, handling of HTTP status codes, and storage of two files (~/.kaggle-agent-id and ~/.kaggle-agent-api-key). The skill explicitly warns not to modify other files and not to leak the API key. There are no instructions to read arbitrary user files, fetch code from unknown URLs, or transmit data to third-party endpoints outside Kaggle.
Install Mechanism
This is an instruction-only skill with no install spec and no code files to drop on disk, which minimizes install-time risk.
Credentials
No environment variables or unrelated credentials are requested. The only persistence is two per-user files in the home directory for agentId and apiKey, which is proportionate for a system that must authenticate to Kaggle. The SKILL.md also instructs secure file permissions (chmod 600) and not to log keys.
Persistence & Privilege
The skill does not request always:true or other elevated persistence. It only instructs creating two credential files under the user's home directory and does not attempt to change system-wide settings or other skills' configs.
Assessment
This skill appears to do what it says: register an agent with Kaggle, save an agentId and a one-time apiToken to two files in your home directory, and interact with Kaggle's API to run the exam. Before installing/running it: (1) Confirm you consent to having an identity registered on Kaggle and to sending exam answers to Kaggle. (2) Be aware the skill will store an API token at ~/.kaggle-agent-api-key — use a dedicated or ephemeral token/account if you prefer to limit exposure. (3) Verify the domain in any prompts or requests is exactly www.kaggle.com as the instructions insist. (4) The skill asks you to include your model identifier when registering; consider whether you are comfortable disclosing that string to Kaggle. (5) If you see the skill ask to read or write files other than the two specified, stop and do not proceed. Overall the components are proportionate to the stated purpose.Like a lobster shell, security has layers — review code before you run it.
latestvk97dveanwdz1qyachfp6zqbteh83zn17
License
MIT-0
Free to use, modify, and redistribute. No attribution required.