The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill generates GitHub Actions workflow files that reference several external actions for security scanning. These include tools from well-known and reputable organizations such as GitHub (actions/checkout, codeql-action), Semgrep (semgrep-action), Aqua Security (trivy-action), and Gitleaks (gitleaks-action). All references use specific version tags, which is a security best practice for CI/CD pipelines.
[INDIRECT_PROMPT_INJECTION]: The skill's operation involves inspecting local project files (like package.json or Dockerfile) to identify the development ecosystem and select appropriate pipeline stages.
Ingestion points: File system inspection of repository root files (e.g., SKILL.md).
Boundary markers: None explicitly defined for file content analysis.
Capability inventory: The skill outputs text (YAML configuration) but does not perform file writes, network requests, or command execution itself.
Sanitization: Not applicable as the skill primarily uses the presence of files rather than their content to trigger template selection.
The risk is minimal as the untrusted data (file names/presence) is used to select pre-defined templates for static configuration generation.

devsecops-pipeline