The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external content from local directories and Git repositories. This inherent risk is mitigated by a mandatory human-in-the-loop approval gate defined in references/deployment-plan-template.md, which requires users to explicitly validate and approve the deployment plan before any resources are provisioned.
Ingestion points: SKILL.md (Stage 0
Resolve repo)
Boundary markers: Mandatory deployment-plan.md review gate requiring status: validated and approval: granted with evidence.
Capability inventory: Volcengine ve CLI, kubectl, docker, ssh (via references/ecs-deploy-steps.md and references/vke-deploy-steps.md).
Sanitization: Automated resolution of environment variables and structured manifest generation.
[COMMAND_EXECUTION]: The skill utilizes the Volcengine ve CLI, docker, and specialized bash scripts (scripts/gen-systemd-unit.sh, scripts/poll-status.sh) to manage cloud infrastructure and application lifecycle. This behavior is consistent with the skill's primary purpose of cloud deployment and follows standard administrative patterns.
[EXTERNAL_DOWNLOADS]: Infrastructure and application components are retrieved from trusted sources, including official Git repositories and well-known container registries such as Docker Hub and Volcengine CR. These operations are documented and integral to the deployment workflow.
[CREDENTIALS_UNSAFE]: The scripts/gen-docker-compose-test.sh file contains hardcoded placeholder passwords (e.g., testpw) for local testing environments. These are used for development simulation purposes in the generated docker-compose.test.yml and do not involve production secrets or sensitive user credentials.

volcengine-deploy