Install
$ agentstack add skill-adibirzu-oci-skills-oci-terraform-authoring ✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
About
OCI Terraform authoring
Create reviewable infrastructure artifacts without contacting OCI. Contact OCI only for discovery, planning, or an explicitly approved apply/destroy after a named-context preflight.
Workflow
- Establish the owner. Default durable resources to
terraform; never give the same resource to direct CLI and Terraform. - Scaffold with
../../scripts/oci_tf.sh scaffold --nameimmediately; it is a safe offline action. - Resolve resource fields from
terraform providers schema -jsonafterterraform init, or from the current official OCI provider docs. Never invent a field. - Keep credentials out of HCL and variable files. Use provider config, environment variables, workload/resource principals, or Vault references.
- Validate with
oci_tf.sh validateand run.tftest.hcltests with mocked providers where possible. If a local prerequisite is missing, deliver the scaffold and the exact next local command rather than stopping the workflow. - Preflight the named context only before creating a binary reviewed plan with
oci_tf.sh plan. - Inspect the metadata-only plan summary. Stop on unexpected replacement/deletion, public exposure, or secret-bearing resources.
- Apply the unchanged plan with
oci_tf.sh apply; usedestroyonly with a separately reviewed destroy plan and destructive approval. - Verify resource state and reconcile any prior CLI break-glass change.
- For team state, prefer the native OCI Object Storage backend when the
reviewed runtime supports it; treat the legacy S3-compatible backend as a deprecated fallback and never place credentials in backend HCL.
Read [terraform-authoring.md](../../references/terraform-authoring.md) for command contracts, provider-schema grounding, discovery, packaging, and state rules.
Common multi-step flows
| Request | Sequence | |---|---| | Write Terraform | intake → scaffold → provider schema/docs → HCL + tests → fmt/init/validate | | Discover existing resources | preflight → empty destination → discover → review generated HCL → remove secrets → choose import/ownership | | Deploy locally | preflight → validate → plan → inspect summary → exact-plan apply → service verification | | Deploy with Resource Manager | author here → package source + schema.yaml → hand off stack/job operations to oci-resource-manager | | Adopt existing resources | prove inventory/ownership → author resource and declarative import blocks → reviewed plan/import → verify zero unintended change | | Refactor addresses | retain moved blocks → plan for no replacement → review consumers → exact-plan apply → verify state addresses | | Investigate drift | refresh-backed plan → classify change → update HCL or restore through owner → reviewed plan → verify convergence | | Configure team state | prefer native OCI backend → verify auth/protection/concurrency → migrate once → test recovery without exposing state | | Destroy | refreshed destroy plan → dependency review → exact destructive approval → apply reviewed plan → verify absence |
Safety boundaries
- Missing context, preflight, live credentials, or approval never blocks offline authoring, tests, validation, review, or documentation; it blocks only the later live operation that requires it.
- Never print or commit state, plan binaries,
.terraform/, wallets, private keys, orterraform.tfvars. - Reject symlinked or non-empty discovery destinations.
- A plan review sidecar is context- and content-bound; changed bytes require a new review.
- Resource discovery is a starting point, not migration proof.
- State can contain secrets even when values are marked sensitive. Encrypt and
access-control it, serialize writers, and never log or attach it.
- Match authentication to the execution context: named local profile or
short-lived token, instance/resource principal, OKE workload identity, or Resource Manager. Never embed key material in provider/backend configuration.
- Pin Terraform, provider, and module constraints; commit reviewed lock metadata;
review module provenance, checksums, licenses, nested dependencies, provisioners, public exposure, and upgrade/migration notes.
- Resolve realm/region endpoints from the installed provider and current
official docs. Do not hardcode commercial-realm domains; use the applicable FIPS-compatible provider where Oracle requires it.
- CLI may inspect or recover. A CLI mutation against a Terraform-owned resource is break-glass and must be followed by
terraform planreconciliation.
Expected output
Report artifact paths, source/schema used, validation result, plan action counts, approval state, verification, and rollback. Never reproduce state or sensitive planned values.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: adibirzu
- Source: adibirzu/oci-skills
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet — be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.