The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the ibmcloud CLI with the ce plugin to manage cloud workloads. The commands are standard for the platform and are used to deploy applications, run jobs, and manage project resources.
[EXTERNAL_DOWNLOADS]: The deployment patterns include Dockerfile templates that pull base images from official and trusted registries, such as the Red Hat Ecosystem Catalog (registry.access.redhat.com), Google Container Registry (gcr.io), and Project Quay (quay.io).
[PROMPT_INJECTION]: The skill analyzes local project metadata (e.g., package.json, requirements.txt, Dockerfile) to provide deployment recommendations. This creates an indirect prompt injection surface where a malicious project could attempt to influence the agent's output.
Ingestion points: Local source code analysis in Phase 1 (ls -la and reading configuration files).
Boundary markers: Not explicitly defined in the file analysis instructions.
Capability inventory: Resource management via ibmcloud ce commands, including creating secrets and environment variables.
Sanitization: The risk is mitigated by the skill's interactive design, which requires the agent to present recommendations and await user confirmation before executing any deployment actions.
[CREDENTIALS_UNSAFE]: The skill demonstrates safe credential management by explicitly instructing the user to store sensitive data like API keys and database URLs in Code Engine secrets rather than including them in plaintext configuration or source code.

code-engine-specialist