The Agent Skills Directory

REMOTE_CODE_EXECUTION (HIGH): The skill documentation and templates include piped shell execution from untrusted sources, which is a critical security risk. Specifically, scripts/scan-image.ts suggests installing Trivy via curl -sfL ... | sh and assets/devcontainer-dockerfile.md contains a template command for installing Poetry using curl -sSL ... | python3 -. Neither aquasecurity nor python-poetry are on the specific list of trusted repositories.
CREDENTIALS_UNSAFE (HIGH): Multiple configuration templates in assets/devcontainer-compose.md contain hardcoded default credentials for PostgreSQL and MySQL services (e.g., POSTGRES_PASSWORD: postgres), which could be inadvertently deployed.
COMMAND_EXECUTION (MEDIUM): The scripts/scan-image.ts script uses the Deno.Command API to execute external binaries (trivy), creating a significant attack surface.
PROMPT_INJECTION (LOW): The script scripts/scan-image.ts is vulnerable to indirect prompt injection through command injection via unsanitized CLI arguments. 1. Ingestion points: imageName variable derived from Deno.args in scripts/scan-image.ts. 2. Boundary markers: Absent. 3. Capability inventory: Subprocess execution via Deno.Command in scripts/scan-image.ts. 4. Sanitization: Absent; the script only filters out arguments starting with dashes.

devcontainer