The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes security alerts and log data from Elasticsearch, which are external data sources that could be influenced by an attacker to include malicious instructions.
Ingestion points: fetch-next-alert.js retrieves data from the .alerts-security.alerts-* index, and run-query.js retrieves data from logs-* patterns.
Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands when the agent processes alert content.
Capability inventory: The skill possesses the capability to execute shell scripts, write temporary query files to the workspace, and perform authenticated network requests to Elasticsearch and Kibana APIs.
Sanitization: The provided scripts format the data for display (e.g., in formatAlertSummary) but do not implement sanitization or filtering to prevent the agent from interpreting alert content as instructions.
[COMMAND_EXECUTION]: The skill is designed around the execution of local Node.js scripts (fetch-next-alert.js, run-query.js, acknowledge-alert.js) via shell commands to perform its primary functions. It also utilizes shell redirection (echo "..." > query.esql) to manage query files.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the @elastic/elasticsearch client and other dependencies via npm install during the setup phase.
[SAFE]: The network communication is directed to user-configured Elasticsearch and Kibana instances defined in environment variables. The scripts include standard security practices for SOC workflows, such as confirmation prompts (--yes required for auto-execution) and dry-run modes for bulk actions. A configuration option allows disabling TLS certificate verification (KIBANA_INSECURE), which is a known security trade-off for internal environments.

security-alert-triage