The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides access to the panda CLI binary and an execute_python tool, which allow the agent to run Python code and shell commands for data analysis. These operations are performed within a workspace environment to facilitate processing blockchain and metric data.
[EXTERNAL_DOWNLOADS]: Fetches data from external blockchain infrastructure, including ClickHouse for blockchain events, Prometheus for infrastructure metrics, Loki for log data, and various Dora explorer API endpoints. These interactions are fundamental to the skill's purpose of network investigation.
[DATA_EXFILTRATION]: Features a storage.upload functionality that allows local files (such as generated charts or data summaries) to be uploaded to an external storage service. This service provides a public URL for the artifacts, which is a core feature for sharing visualization results with users.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it retrieves data from external sources (logs, events, and API responses) that could contain malicious strings intended to influence the agent's behavior.
Ingestion points: ClickHouse database records, Loki logs, Prometheus metric labels, and Dora API response bodies.
Boundary markers: The instructions do not implement specific delimiters or 'ignore embedded instructions' warnings when processing retrieved data.
Capability inventory: The skill has access to Python execution (execute_python) and file system writes in the /workspace/ directory.
Sanitization: There is no evidence of sanitization, filtering, or validation performed on external data strings before they are passed into the agent's context for analysis.

query