The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous scripts and configurations that execute shell commands for security automation. Examples include:
Execution of security scanning tools (trufflehog, semgrep, snyk, checkov, trivy, zap) within CI/CD pipelines.
Automation of Kubernetes deployment and configuration (kubectl apply, aws eks update-kubeconfig).
Local script execution for container scanning (scripts/scan-container.sh) and vulnerability processing (scripts/process-vulnerabilities.py).
[EXTERNAL_DOWNLOADS]: The skill references and downloads tools and configurations from well-known and trusted technology services and organizations, including:
GitHub Actions from trusted repositories like trufflesecurity/trufflehog, aquasecurity/trivy-action, bridgecrewio/checkov-action, and returntocorp/semgrep-action.
Kubernetes configuration manifests from github.com/bitnami-labs/sealed-secrets.
Docker images from ghcr.io and docker.io (e.g., hashicorp/vault, aquasec/trivy).
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8c: Tool output poisoning).
Ingestion points: The scripts scripts/process-vulnerabilities.py and scripts/create-vulnerability-tickets.py ingest JSON data from dependency and container scanners (Snyk, Trivy). This data originates from external vulnerability databases (CVE/CWE entries) that an attacker could potentially influence.
Boundary markers: There are no explicit boundary markers or warnings to ignore embedded instructions in the processed data.
Capability inventory: The skill possesses capabilities to write to Jira (via atlassian/gajira-create), post to Slack, and influence CI/CD pipeline outcomes. The scripts have access to environment-stored secrets like JIRA_API_TOKEN and GITHUB_TOKEN.
Sanitization: The Python scripts interpolate external descriptions directly into Markdown and HTML outputs without sufficient escaping or sanitization to prevent the injection of malicious instructions or payload rendering.

devsecops