The Agent Skills Directory

[SAFE]: The skill instructions and reference materials promote high-security standards for Docker usage.
Recommends running processes as a non-root user and explicit USER definitions.
Advises against using --privileged flags and unnecessary capabilities (cap_drop).
Encourages using BuildKit secrets (--mount=type=secret) to avoid credential exposure in image layers.
Promotes the use of official vulnerability scanners like trivy, grype, and docker scout to audit images before publication.
References official Docker documentation and established OCI standards.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted configuration files, which presents a surface for indirect injection.
Ingestion points: Processes user-provided Dockerfile, docker-compose.yml, and .dockerignore files as part of its core workflow (SKILL.md).
Boundary markers: The instructions do not define specific delimiters or "ignore previous instructions" guards for the content of these files.
Capability inventory: The skill suggests performing builds and scans, implying access to shell or terminal tools (docker build, trivy, etc.).
Sanitization: There is no mention of sanitizing or escaping the content of processed files before they are interpreted by the agent.
Note: This is considered SAFE in this context as the skill's primary purpose is to audit these files, and it includes robust instructions to reject unsafe patterns found within them.

specsfy-specialist-docker