# Send tf-plan-review to your agent Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually. ## Fast path - Download the package from Yavira. - Extract it into a folder your agent can access. - Paste one of the prompts below and point your agent at the extracted folder. ## Suggested prompts ### New install ```text I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Then review README.md for any prerequisites, environment setup, or post-install checks. Tell me what you changed and call out any manual steps you could not complete. ``` ### Upgrade existing ```text I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Then review README.md for any prerequisites, environment setup, or post-install checks. Summarize what changed and any follow-up checks I should run. ``` ## Machine-readable fields ```json { "schemaVersion": "1.0", "item": { "slug": "tf-plan-review", "name": "tf-plan-review", "source": "tencent", "type": "skill", "category": "数据分析", "sourceUrl": "https://clawhub.ai/tkuehnl/tf-plan-review", "canonicalUrl": "https://clawhub.ai/tkuehnl/tf-plan-review", "targetPlatform": "OpenClaw" }, "install": { "downloadUrl": "/downloads/tf-plan-review", "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=tf-plan-review", "sourcePlatform": "tencent", "targetPlatform": "OpenClaw", "packageFormat": "ZIP package", "primaryDoc": "SKILL.md", "includedAssets": [ "CHANGELOG.md", "README.md", "SECURITY.md", "SKILL.md", "TESTING.md", "scripts/tf-plan-review.sh" ], "downloadMode": "redirect", "sourceHealth": { "source": "tencent", "slug": "tf-plan-review", "status": "healthy", "reason": "direct_download_ok", "recommendedAction": "download", "checkedAt": "2026-04-30T17:18:17.996Z", "expiresAt": "2026-05-07T17:18:17.996Z", "httpStatus": 200, "finalUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=tf-plan-review", "contentType": "application/zip", "probeMethod": "head", "details": { "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=tf-plan-review", "contentDisposition": "attachment; filename=\"tf-plan-review-0.2.1.zip\"", "redirectLocation": null, "bodySnippet": null, "slug": "tf-plan-review" }, "scope": "item", "summary": "Item download looks usable.", "detail": "Yavira can redirect you to the upstream package for this item.", "primaryActionLabel": "Download for OpenClaw", "primaryActionHref": "/downloads/tf-plan-review" }, "validation": { "installChecklist": [ "Use the Yavira download entry.", "Review SKILL.md after the package is downloaded.", "Confirm the extracted package contains the expected setup assets." ], "postInstallChecks": [ "Confirm the extracted package includes the expected docs or setup files.", "Validate the skill or prompts are available in your target agent workspace.", "Capture any manual follow-up steps the agent could not complete." ] } }, "links": { "detailUrl": "https://openagent3.xyz/skills/tf-plan-review", "downloadUrl": "https://openagent3.xyz/downloads/tf-plan-review", "agentUrl": "https://openagent3.xyz/skills/tf-plan-review/agent", "manifestUrl": "https://openagent3.xyz/skills/tf-plan-review/agent.json", "briefUrl": "https://openagent3.xyz/skills/tf-plan-review/agent.md" } } ``` ## Documentation ### Terraform Plan Analyzer & Risk Assessor Analyze terraform plan output and produce an AI-powered risk assessment of every infrastructure change — before you press apply. This skill is STRICTLY READ-ONLY. It runs terraform plan and terraform validate to analyze changes, but it NEVER runs terraform apply, terraform destroy, terraform import, terraform taint, or any command that modifies infrastructure or state. ### Activation This skill activates when the user mentions: "terraform plan", "tf plan", "review plan", "plan review" "is this plan safe", "safe to apply", "risk assessment" "what will be destroyed", "what changes", "terraform changes" "terraform state", "state drift", "drift detection" "terraform validate", "validate config", "tf validate" "IAM changes", "security group changes", "infrastructure changes" "blast radius", "cascade effects", "dependencies" "tofu plan", "opentofu" (same workflow, different binary) ### Example Prompts "Review this terraform plan before I apply" "What will be destroyed in this plan?" "Is this plan safe to apply?" "Show me the state drift" "What IAM changes are in this plan?" "Validate my terraform config in ~/infra/prod" "Run a risk assessment on the terraform plan in /deployments/staging" "What's the blast radius if I apply this plan?" ### Permissions permissions: exec: true # Required to run terraform/tofu CLI read: true # Read .tf files and plan output write: false # NEVER writes — strictly read-only analysis network: true # terraform plan needs provider API access ### Terraform Change Types — What the Agent Must Know Understanding Terraform change types is critical for accurate risk assessment: ### Action Types (from plan JSON) ActionMeaningRisk ProfilecreateNew resource being addedGenerally safe (unless IAM/security)updateExisting resource modified in-placeModerate (depends on what's changing)deleteResource being permanently destroyedDANGEROUS — data loss riskreplace (delete + create)Resource must be destroyed and recreatedDANGEROUS — downtime + data lossreadData source being refreshedSafe (read-only)no-opNo changes neededSafe ### What Makes a Change Dangerous Critical (🔴 CRITICAL): Any destroy/replace of: IAM roles/policies, security groups, KMS keys, secrets, databases (RDS, DynamoDB, Cloud SQL, Azure SQL), S3 buckets, DNS records, WAF rules, CloudTrail Any update to IAM policies, security group rules, encryption settings These changes can cause data loss, security breaches, or service outages Dangerous (🟠 DANGEROUS): Destroy/replace of: EC2 instances, load balancers, ECS/EKS clusters, VPCs, subnets, NAT gateways, Lambda functions, API gateways These changes cause downtime and may require manual intervention to recover Moderate (🟡 MODERATE): Updates to: autoscaling policies, monitoring/alerting rules, launch templates Creates of: security-sensitive resources (new IAM roles, new security groups) Changes that affect capacity or observability but not data integrity Safe (🟢 SAFE): Tag-only updates Creating new non-sensitive resources No-op / read operations ### Replace is Especially Dangerous When Terraform says it must "replace" a resource, it means: Delete the existing resource (irreversible) Create a new one with the new configuration This is triggered when an immutable attribute changes (e.g., changing RDS engine_version, EC2 ami, changing a subnet's AZ). The agent should always flag replaces prominently because: The old resource (and its data) is destroyed There will be a gap between destroy and create (downtime) Dependent resources may break during the transition ### Agent Workflow Follow this sequence exactly based on user intent: ### For Plan Analysis ("review this plan", "is it safe", "what changes") Step 1: Run Plan Analysis bash /scripts/tf-plan-review.sh plan If no directory specified, use the current working directory. The script outputs: stdout: Structured JSON with all resource changes, risk classifications, and summary stderr: Beautiful Markdown risk report Step 2: Interpret the JSON Parse the JSON output. Key fields: { "overall_risk": "🔴 CRITICAL | 🔴 HIGH | 🟡 MODERATE | 🟢 LOW", "summary": { "create": 5, "update": 3, "destroy": 1, "replace": 0 }, "risk_breakdown": { "critical": 1, "dangerous": 0, "moderate": 2, "safe": 5 }, "resources": [ { "address": "aws_iam_role.admin", "action": "delete", "risk": "🔴 CRITICAL" } ] } Step 3: Present the Risk Assessment Show the Markdown report from stderr. Then add your own AI analysis: Lead with the overall risk level — make it viscerally clear Highlight destroys and critical changes first — these are what kill production Explain WHY each critical change is dangerous in plain English Assess blast radius — what other resources depend on the destroyed ones? Present the pre-apply checklist — what should the human verify? Give a clear recommendation: "Safe to apply" / "Review needed" / "DO NOT APPLY without ___" Tone guidance for critical plans: Don't be polite about danger. If a plan destroys a production database, say so bluntly. "This plan will permanently delete your RDS instance prod-db. All data will be lost. Do you have a backup?" Make the "oh shit" moment impossible to miss. ### For State Inspection ("show me state", "what's managed", "state drift") bash /scripts/tf-plan-review.sh state "" The filter is optional — it greps resource addresses. Examples: bash /scripts/tf-plan-review.sh state "iam" . → all IAM resources bash /scripts/tf-plan-review.sh state "aws_instance" . → all EC2 instances bash /scripts/tf-plan-review.sh state "" . → all resources ### For Validation ("validate config", "check syntax") bash /scripts/tf-plan-review.sh validate Reports configuration errors and warnings without running a plan. ### Environment Variables VariableDefaultDescriptionTF_BINARYauto-detectOverride binary: terraform, tofu, or a pathTF_PLAN_TIMEOUT600Timeout for terraform plan in seconds The script auto-detects terraform first, then tofu. Set TF_BINARY=tofu to force OpenTofu. ### Error Handling SituationBehaviorterraform/tofu not foundJSON error with install links for bothjq not foundJSON error with install linkNo .tf files in directoryJSON error: "No Terraform configuration files found"Not initializedAuto-runs terraform init (for plan) or terraform init -backend=false (for validate)Plan fails (provider errors)Extracts error from plan JSON diagnostics, reports itPlan timeoutProcess killed after TF_PLAN_TIMEOUT secondsState not foundJSON error explaining no state existsEmpty stateReports "State is empty — no managed resources" ### Safety — CRITICAL RULES NEVER run terraform apply — not even with -auto-approve, not even with -target, not even "just this one resource". NEVER. NEVER run terraform destroy — not under any circumstances. NEVER run terraform import — this modifies state. NEVER run terraform taint or terraform untaint — these modify state. NEVER run terraform state mv, terraform state rm, or terraform state push — these modify state. Never expose cloud credentials — if they appear in plan output, redact them. Handle sensitive values — Terraform marks values as (sensitive). Never try to reveal them. Never cache or store plan output — plans can contain secrets in resource attributes. The ONLY terraform commands this skill runs are: plan, show, state list, state show, validate, init, providers. If the user asks you to apply a plan, respond: "I can analyze and assess Terraform plans, but I cannot apply them. Applying infrastructure changes requires human review and explicit execution. Based on my analysis, here's what you should verify before running terraform apply..." ### "Is this plan safe to apply?" Run the plan analysis. If overall_risk is 🟢 LOW: "This plan looks safe. It creates X new resources with no destroys or security changes. The pre-apply checklist is straightforward." If overall_risk is 🔴 CRITICAL: "⚠️ This plan has CRITICAL risk. [Explain specific dangers]. I strongly recommend review by another team member before applying." ### "What will be destroyed?" Run plan, then filter for action == "delete" or action == "replace". Present each with: Resource address Resource type Why it matters (is it stateful? does it have data?) What depends on it ### "What IAM changes are in this plan?" Run plan, then filter resources matching IAM patterns. For each: What permission is changing Is it adding or removing access Is it overly permissive (e.g., Action: *) ### "Show me the blast radius" Run plan, identify all destroys/replaces, then explain: What other resources reference the destroyed ones What will break when the resource is gone Whether Terraform will auto-fix the dependencies or if manual intervention is needed ### Discord v2 Delivery Mode (OpenClaw v2026.2.14+) When the conversation is happening in a Discord channel: Send a compact first summary (overall risk, destroy count, critical resources), then ask if the user wants the full report. Keep the first response under ~1200 characters and avoid large Markdown tables in the first message. If Discord components are available, include quick actions: Show Critical Changes Show Destroyed Resources Show Pre-Apply Checklist If components are not available, provide the same follow-ups as a numbered list. Prefer short follow-up chunks (<=15 lines per message) for large plans. ### Sensitive Data Handling Terraform plan JSON may contain sensitive values. The script does NOT extract resource attribute values — it only extracts resource addresses, types, and actions. However, when presenting results: Never show attribute values marked (sensitive) by Terraform Never show provider credentials or backend configuration secrets If a user asks "what value is changing?", explain that you can see the change type but sensitive values are redacted by Terraform for security Never store or cache plan output files ## Trust - Source: tencent - Verification: Indexed source record - Publisher: tkuehnl - Version: 0.2.1 ## Source health - Status: healthy - Item download looks usable. - Yavira can redirect you to the upstream package for this item. - Health scope: item - Reason: direct_download_ok - Checked at: 2026-04-30T17:18:17.996Z - Expires at: 2026-05-07T17:18:17.996Z - Recommended action: Download for OpenClaw ## Links - [Detail page](https://openagent3.xyz/skills/tf-plan-review) - [Send to Agent page](https://openagent3.xyz/skills/tf-plan-review/agent) - [JSON manifest](https://openagent3.xyz/skills/tf-plan-review/agent.json) - [Markdown brief](https://openagent3.xyz/skills/tf-plan-review/agent.md) - [Download page](https://openagent3.xyz/downloads/tf-plan-review)