The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to interact with production environments by executing platform-specific CLI commands, including vercel, netlify, wrangler, docker, and kubectl for building and deploying applications.
[INDIRECT_PROMPT_INJECTION]: The skill reads project-level configuration files (such as vercel.json, netlify.toml, and Dockerfile) to determine deployment strategies, creating a potential surface for indirect prompt injection from untrusted codebases.
Ingestion points: Workflow Step 1 in SKILL.md instructs the agent to read platform signals from various local configuration and manifest files.
Boundary markers: The skill uses a structured 'Output contract' and explicit workflow steps to maintain control over the deployment process.
Capability inventory: The agent has the capability to execute shell commands for builds, deployments, and rollbacks, as well as perform network requests for health checks and smoke tests.
Sanitization: The skill explicitly mandates that secret values must never be printed or echoed and should only be managed through the platform's native environment variable mechanisms.

runway