The Agent Skills Directory

[PROMPT_INJECTION]: The scripts log_analyzer.py and health_check_validator.py ingest data from external sources such as log files and HTTP response bodies. This creates a surface for indirect prompt injection, where an attacker might attempt to influence the agent's behavior by embedding malicious instructions in the data being monitored. This risk is characteristic of observability tools and is not indicative of malicious intent.\n
Ingestion points: scripts/log_analyzer.py (reads log lines), scripts/health_check_validator.py (fetches response bodies).\n
Boundary markers: None identified in the processing logic.\n
Capability inventory: The skill has network access and file read capabilities via standard libraries.\n
Sanitization: Scripts perform basic regex matching and status code validation but do not implement specific defenses against prompt injection in monitored content.\n- [EXTERNAL_DOWNLOADS]: Documentation in references/tool_comparison.md provides an example installation command for the Datadog agent from a well-known industry source (https://s3.amazonaws.com/dd-agent/scripts/install_script.sh). This reference is used for illustrative purposes and points to a trusted provider.\n- [SAFE]: No hardcoded credentials, malicious persistence mechanisms, or unauthorized privilege escalation patterns were found. The skill's behavior and dependencies are appropriate for its described use case in system administration and monitoring.

monitoring-observability