- Home
- Skills
- DevOps & Cloud
- terraform boundary
terraform boundary
Enforce human-in-the-loop cloud deployment and security gates for Agentic-AI-Systems infrastructure.
Free
Works with the AI Agents you already use
terraform boundary
Example session with this skill installed
I need to add a new read-only weather tool to my AWS agent infrastructure in the dev environment. Fetch the tree and prepare the change.
- Read your context and instructions
- Compiled the terraform boundary
I have fetched the AWS infrastructure tree and initialized the boundary checks. The environment is clean and ready for the new tool definition.
Infrastructure Initialization
| Action | Result |
|---|---|
| Fetch Tree | terraform-aws/ successfully created from pinned release v0.1.0 |
| Pillar 1 Check | Guard hook verified. Lock 1 active. |
| Initial Gate | boundary.sh check passed. 0 violations found. |
| Target File | terraform-aws/envs/dev/main.tf |
Security Review (Draft)
- Compromised orchestrator: No new write access. The tool is labeled
read. - Prompt-injected model: No path to write tools created.
- Leaked approval claim: N/A (Read-only tool).
- Terraform change: AWS identity policy remains restricted to
ReadOnlyAccess.
Next steps
- Add the tool ARN to
tool_function_arnsin yourdevenvironment. - Run
scripts/boundary.sh checkto verify the new configuration. - Use the generated evidence record to proceed to human approval.
Connects securely to your tools. The creator never sees your data.
About this skill
Standard infrastructure as code often lacks the strict guardrails needed for agentic systems. This skill enforces a human-in-the-loop deployment boundary for AI infrastructure built on the Agentic-AI-Systems framework. It prevents agents from modifying their own cloud resources while providing a verifiable audit trail for regulated environments.
What it does
- Enforce human-only execution by blocking state-changing Terraform commands within the agent's environment.
- Implement security gates that validate every infrastructure change against a predefined threat model.
- Verify cloud boundaries across AWS, Azure, GCP, and Snowflake to ensure write tools are unreachable by the agent orchestrator.
- Generate evidence records including security reviews and compliance checklists for regulated deployments.
- Pin infrastructure versions to a specific release of Agentic-AI-Systems to prevent configuration drift.
How it works
- Initialize the tree by fetching a specific cloud infrastructure template and running initial safety checks.
- Modify configuration within local environment directories to add tools or adjust resource sizing.
- Run security gates using the
boundary.sh checkcommand to verify that no security properties were compromised by edits. - Draft evidence by completing the security review and compliance checklists required for the audit log.
- Preview and handoff by generating a plan with read-only credentials and providing the exact command for a human to apply.
Frameworks & tools
This skill uses Terraform and Open Policy Agent (OPA). It integrates with AWS, Azure, GCP, and Snowflake. It requires conftest for policy enforcement and python3 for the guard hook.
Why this beats prompting it yourself
Standard prompts cannot stop an agent from executing a shell command if the model ignores the instruction. This skill uses a Python-based guard hook that intercepts and blocks unauthorized binaries at the process level. It moves security from a "suggestion" to an identity-enforced property of the cloud platform.
Use cases
- Deploying a production AI agent in a regulated industry like finance or healthcare.
- Implementing a "least privilege" architecture for an agent that needs to call external APIs.
- Creating an audit trail for infrastructure changes performed by an AI assistant.
- Testing new agent tools in a sandbox without risking account-wide credentials.
Known limitations
Does not support remote state backends, .tf.json files, or Terraform provisioners. Requires a local installation of the guard.py hook to function.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for DevOps & Cloud.
Requires local Python guard hook installation. No support for remote state, symlinks, or Terraform provisioners. Limited to Agentic-AI-Systems v0.1.0 tree.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 21 days ago
- Free with an Agensi account
- Accounts
GitHub
Conftest
Hashicorp
Permissions
Allowed Hosts