The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill manages Volcengine access keys and secret keys provided via environment variables (VOLCENGINE_ACCESS_KEY, VOLCENGINE_SECRET_KEY). It demonstrates strong security posture by:
Mapping these credentials to standard AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY variables exclusively for the purpose of authorizing the Terraform S3-compatible backend (Volcengine TOS).
Explicitly instructing users in SKILL.md and references/backend-tos.md to avoid committing state files (.tfstate) and to add sensitive outputs to .gitignore.
Utilizing helper scripts (scripts/export_outputs.sh) that automatically set file mode 0600 on generated JSON output files containing sensitive data like kubeconfigs or database credentials.
Providing a dedicated "Safety Rules" section in SKILL.md that warns against reading plaintext credentials from local config files.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface in scripts/gen_tfvars.py, which reads and processes a deployment context JSON file (.volcengine/deploy-choice.json). This could potentially be used for indirect injection if a malicious source influences the JSON content. However, the risk is mitigated as the script includes a hcl_string function that properly sanitizes values by escaping double quotes and backslashes before writing them to the terraform.tfvars file.
[DYNAMIC_EXECUTION]: The skill employs dynamic generation and shell execution to support its primary infrastructure tasks:
scripts/gen_tfvars.py generates HCL variable files based on dynamic input.
Various Terraform examples (e.g., assets/examples/volcengine-vke-cr-nginx/workload/main.tf) use the local-exec provisioner to run Docker commands for logging into private registries and pushing images. These commands are fixed, use validated variables, and are transparently documented in the accompanying reference files.

volcengine-iac