The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes workflows where the agent ingests and analyzes external data, including trace outputs and LLM responses from the applications being monitored. This creates a surface area for indirect prompt injection where instructions embedded in the analyzed data could influence the agent's behavior.
Ingestion points: The agent is instructed to read trace data, span inputs/outputs, and evaluation results from the Laminar platform during debugging and fix-and-verify loops (references/debugging.md, references/eval-loop.md).
Boundary markers: The skill emphasizes the use of privacy controls and redaction functions (ignore_inputs, ignore_output) to prevent sensitive data from entering the trace context, which serves as a partial boundary for data content.
Capability inventory: The skill utilizes lmnr-cli to perform project configuration, SQL querying of telemetry data, and dataset management. These operations involve network requests to the Laminar API and local file modifications.
Sanitization: The instructions provide detailed guidance on redacting PII and secrets using SDK-level privacy controls, which mitigates the risk of exfiltrating sensitive data to the observability platform.
[COMMAND_EXECUTION]: The skill makes extensive use of the lmnr-cli tool to manage the observability lifecycle. Commands include setup (which modifies local configuration files and installs agent skills), login (authenticating via OAuth), sql query (retrieving telemetry), and debug session management.
[EXTERNAL_DOWNLOADS]: The skill references the installation and use of official vendor packages, including the lmnr Python package (with optional extras like lmnr[all]) and Node.js packages @lmnr-ai/lmnr and lmnr-cli.
[DATA_EXFILTRATION]: The core functionality of the skill involves transmitting application traces, spans, and metadata to the Laminar platform (api.lmnr.ai or a user-defined self-hosted endpoint). The documentation explicitly instructs users to manage LMNR_PROJECT_API_KEY via environment variables and provides privacy mechanisms to prevent the inadvertent capture of sensitive data.

laminar