The Agent Skills Directory

[DATA_EXFILTRATION]: The script scripts/ask_teams_agent.py extracts authentication tokens (the deepexi_auth cookie and localStorage session tokens) from the local DeepWorks desktop application by connecting to its Chrome DevTools debug ports (9222, 9223, 9228). These credentials are sent to the vendor's cloud endpoints at test.deepworks.cn and agentteam.deepworks.cn to authenticate API requests.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data that could be used for indirect prompt injection.
Ingestion points: The skill retrieves agent metadata (names and descriptions) from the cloud API and accepts user-provided tasks via the --message argument.
Boundary markers: No specific delimiters or safety instructions are used to separate user input from system prompts in the API payload.
Capability inventory: The script performs network operations, reads configuration files from the local file system, and writes logs to a temporary directory.
Sanitization: No validation or sanitization of the input message content is performed before it is sent to the cloud backend.
[COMMAND_EXECUTION]: The script uses the Python standard library urllib to execute network operations against the Agent Teams cloud API to manage sessions and retrieve task results.

dp-ask-teams