The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and process external data such as threat intelligence feeds and CVE disclosures. While this represents a potential surface for indirect prompt injection, the skill mitigates this by including a 'Prompt Injection Defense Directive' and a mandatory sanitization step (Workflow 2, Step 1) that includes scanning for adversarial patterns.
Command Execution for Local Validation: The skill utilizes a local Python script (scripts/validate_yaral_rule.py) to perform deterministic validation of YARA-L rules. This script relies on the standard library and does not perform network operations or access sensitive system files beyond the input file provided by the user.
Cloud Management and Attribution: The skill provides instructions for executing gcloud commands and REST API calls to manage SecOps rules. It adheres to security best practices by recommending usage attribution through environment variables and specific User-Agent headers, which is a standard practice for enterprise cloud environments.
Ingestion Points (Category 8): External threat intelligence feeds, CVE disclosures, and synthetic UDM events (Workflow 2, Step 1).
Boundary Markers (Category 8): The skill includes a 'Prompt Injection Defense Directive' and a mandatory security gate in the extraction process.
Capability Inventory (Category 8): Tools for rule creation (create_rule), validation (validate_rule), and gcloud CLI commands.
Sanitization (Category 8): Detailed procedure in Workflow 2, Step 1 for stripping HTML, normalizing text, and scanning for injection patterns.

secops-detection-engineering