The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes performance data from external sources that could potentially contain adversarial content designed to influence the agent's behavior.
Ingestion points: Data is ingested from Prometheus, StatsD, and CloudWatch APIs via the scripts/metrics_collection.py script as documented in scripts/README.md.
Boundary markers: No explicit delimiters or specific 'ignore' instructions are present in the core instructions in SKILL.md to protect the agent from processing malicious payloads embedded in metric names or values.
Capability inventory: The skill is granted access to the Write tool for configuration and dashboard generation, and Bash execution capabilities for processing scripts.
Sanitization: The inclusion of scripts/metrics_validation.py for schema and threshold validation serves as a mitigation layer to ensure data quality and integrity.
[COMMAND_EXECUTION]: The skill utilizes restricted shell commands to execute bundled automation scripts.
Evidence: The allowed-tools section in SKILL.md permits the use of Bash(prometheus:*), Bash(metrics:*), and Bash(monitoring:*). These tools are used to run the Python scripts described in scripts/README.md (metrics_collection.py, metrics_aggregation.py, etc.) for managing the monitoring lifecycle.

aggregating-performance-metrics