The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the scanning of external artifacts such as container images and filesystems, which presents a potential surface for indirect prompt injection if those artifacts contain malicious metadata designed to influence the agent.
Ingestion points: The skill uses commands to scan local and remote container images (trivy image), project filesystems (trivy fs), and SBOM files (grype sbom) as shown in SKILL.md.
Boundary markers: The skill includes explicit warnings in the "Limitations" and "CI Integration" sections regarding authorized scope and the use of read-only inventory steps before active assessment.
Capability inventory: The agent is guided to perform file system reads, network operations via cloud CLIs (AWS, Azure, Google Cloud), and the execution of container security tools.
Sanitization: The skill relies on the internal parsing and filtering of the security scanners; no secondary sanitization of scanner output is implemented in the provided documentation.
[EXTERNAL_DOWNLOADS]: The documentation references the installation of security tools from well-known and established sources.
Evidence: References to brew install trivy, docker pull aquasec/trivy, and brew install grype within SKILL.md.
[COMMAND_EXECUTION]: The skill provides templates for executing standard security scanners and cloud-native CLI tools to perform vulnerability assessments and manage repository security policies.
Evidence: Command examples for trivy, grype, docker scout, aws ecr, az acr, and gcloud artifacts are provided in SKILL.md for their intended security purposes.

container-scanning