The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The installation instructions direct the agent or user to clone a repository and execute its content, which is a high-risk pattern when the source is not verified.
Evidence: The installation block in SKILL.md contains git clone https://github.com/waruna834/API_Client___Tester.git followed by python main.py.
[EXTERNAL_DOWNLOADS]: The skill fetches source code from a third-party GitHub repository that is not associated with a trusted organization or the skill author.
Evidence: https://github.com/waruna834/API_Client___Tester.git in the installation section.
[COMMAND_EXECUTION]: The skill relies on shell command execution for its setup process, which can be exploited if the source repository is compromised.
Evidence: Commands include git clone, pip install -r requirements.txt, and python main.py.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external API, GraphQL, and WebSocket endpoints, providing a vulnerability surface for indirect injection attacks.
Ingestion points: Data is ingested from external servers in APIClient.request (via requests.request), GraphQLClient.query (via requests.post), and WebSocketClient.on_message (via websocket.WebSocketApp).
Boundary markers: The skill does not implement delimiters or safety instructions to prevent the agent from interpreting API response data as commands.
Capability inventory: The skill has the ability to perform arbitrary network requests (requests.request), manage file collections (CollectionManager.save_collection), and requires shell access for installation.
Sanitization: No sanitization or validation is performed on the data returned from HTTP responses or WebSocket messages before it is presented to the agent.

api-client-tester-waruna834