The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of conversation turns and simulation messages from the LangWatch platform into the local agent's context. This creates a surface for indirect prompt injection if malicious data is sent from the platform.
Ingestion points: The connector receives conversation messages, newMessages, and session data via WebSocket or HTTP long polling.
Boundary markers: No explicit boundary markers or sanitization instructions are provided for the agent's primary input stream in the connector implementation.
Capability inventory: The agent being connected has access to the codebase's secrets and dependencies. The skill also utilizes the code_access tool for modifications and performs network operations via the SDK.
Sanitization: Local scrubbing is performed by the langwatch report tool, but the agent connector itself does not define sanitization logic for simulation inputs.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to install external packages and tools essential for its operation.
Evidence: Commands provided include pip install langwatch, npm install langwatch zod, and npx --yes langwatch report.
[COMMAND_EXECUTION]: The skill uses various vendor-specific CLI commands to manage agent registration and execute tests.
Evidence: CLI usage includes langwatch agent list, langwatch scenario create, langwatch test-suite run, and langwatch agent tunnel.

connect-agent