The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data such as Terraform plan outputs and GitHub Action workflow files, presenting a standard attack surface for indirect prompt injection.
Ingestion points: scripts/check-action-refs.sh reads content from workflow YAML files; assets/github-actions-terraform.yml reads the content of plan.txt to post as a pull request comment.
Boundary markers: The GitHub Action template in assets/github-actions-terraform.yml wraps processed plan output in markdown code blocks to provide clear delimitation from other prompt instructions.
Capability inventory: The skill utilizes the Bash tool to execute infrastructure commands, file operations, and network queries.
Sanitization: scripts/check-action-refs.sh performs string parsing and validation on extracted action references before resolving them via API calls.
[EXTERNAL_DOWNLOADS]: The skill's CI/CD templates and scripts interact with well-known and official services for configuration and verification.
Fetches official GitHub Actions from repositories such as hashicorp/setup-terraform, aws-actions/configure-aws-credentials, and aquasecurity/trivy-action.
Queries the GitHub API (api.github.com) in scripts/check-action-refs.sh to resolve and verify the existence of action reference tags and commit SHAs.
[COMMAND_EXECUTION]: The skill uses the Bash tool to manage infrastructure state and perform operations.
Executes terraform or tofu binaries for planning, applying changes, and performing state surgery.
Runs utility scripts for staleness verification of workflow references and automated testing.

terraform-ops