The Agent Skills Directory

[COMMAND_EXECUTION]: The Python script scripts/agent.py dynamically constructs Splunk Search Processing Language (SPL) queries using f-string interpolation for parameters like incident_id and ioc_value. This pattern is vulnerable to SPL injection if the input data contains malicious Splunk operators or unbalanced quotes.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external SIEM indices and lookup tables which are untrusted and could contain attacker-controlled payloads designed to manipulate dashboard views or automated logic.
Ingestion points: Untrusted data enters the agent via Splunk search results and CSV lookups in scripts/agent.py and SKILL.md (Steps 2, 3, 4, 7).
Boundary markers: Variables are enclosed in double quotes within SPL queries, but no explicit 'ignore instructions' markers or delimiters are used for the ingested content.
Capability inventory: The skill has the capability to execute searches on a Splunk instance via the splunk-sdk and write data to the local filesystem (ir_dashboard_report.json).
Sanitization: No input validation, escaping, or character sanitization is applied to data before it is interpolated into SPL command strings.

building-incident-response-dashboard