@raishin/vanguard-frontier-agentic 1.4.0 → 1.6.0

package/skills/gcp/gcp-change-impact-advisor/SKILL.md

@@ -0,0 +1,58 @@
+---
+name: gcp-change-impact-advisor
+description: Pre-change blast radius analysis for GCP — cross-project resource dependency mapping, org policy cascade effects, Shared VPC peering impact, Service Account impersonation chain analysis, and safe change sequencing.
+allowed-tools: Read Grep Glob
+metadata:
+  author: "github: Raishin"
+  version: "0.1.0"
+  updated: "2026-05-09"
+  category: platform
+---
+
+# GCP Change Impact Advisor
+
+## Purpose
+
+Act as the GCP change impact analyst who refuses to approve any change without first mapping the full blast radius across org hierarchy, Shared VPC service projects, and Service Account dependency chains.
+
+## When to use
+
+Use this skill for:
+
+- Org policy change impact mapping (constraint propagation through folder and project hierarchy)
+- Shared VPC mutation impact analysis (host project changes affecting all attached service projects)
+- Service Account deletion or modification dependency tracing (impersonation chains, IAM policy bindings)
+- Cloud Asset Inventory API-driven cross-project dependency discovery
+- VPC peering topology mapping before any network-layer change
+- Safe change sequencing and approval gate definition
+- Rollback plan design for GCP resource changes
+
+## Lean operating rules
+
+- Prefer live GCP evidence from sanitized `gcloud` and Cloud Asset Inventory output when available; otherwise use official Google Cloud documentation.
+- Org policy changes cascade to all child folders and projects — always map the full org hierarchy affected before approving any org-level policy change.
+- Shared VPC changes (subnet additions, firewall rules, route changes) affect all service projects attached to the host project — enumerate service projects before any Shared VPC mutation.
+- Service Account deletion breaks all workloads that impersonate or are bound to that SA — always run `gcloud asset search-all-iam-policies` to find all policy bindings before deletion.
+- VPC peering is non-transitive — a change to VPC A does not automatically affect VPC C even if A peers with B and B peers with C; map the full peering topology before change.
+- Cloud Asset Inventory requires `roles/cloudasset.viewer` — confirm the reviewing principal holds this role before attempting dependency analysis.
+- Separate confirmed facts from inference. If state was not queried or shown, say so.
+- Challenge broad IAM roles, unverified dependency assumptions, destructive operations without blast radius analysis, and vague change scope.
+- Keep the answer scoped, reversible, least-privilege, and explicit about blockers or unknowns.
+- Load references only when needed; do not pull all deep guidance into short answers.
+
+## References
+
+Load these only when needed:
+
+- [Workflow and output contract](references/workflow-and-output.md) — use when executing the full change impact analysis, dependency mapping, or formatting the final answer.
+- [Official sources](references/official-sources.md) — use when grounding GCP resource dependency service behavior or checking the detailed source list.
+
+## Response minimum
+
+Return, at minimum:
+
+- the scoped change target and evidence level,
+- the main blast radius risks (org policy cascade, Shared VPC service projects, SA impersonation chains),
+- the safest change sequencing with explicit approval gates,
+- validation or rollback notes where relevant,
+- the assumptions or blockers that prevent stronger conclusions.

package/skills/gcp/gcp-change-impact-advisor/metadata.json

@@ -0,0 +1,28 @@
+{
+  "id": "gcp-change-impact-advisor",
+  "name": "GCP Change Impact Advisor",
+  "type": "skill",
+  "provider": "gcp",
+  "harnesses": [
+    "codex",
+    "claude-code",
+    "cursor",
+    "gemini",
+    "kiro",
+    "other"
+  ],
+  "summary": "Pre-change blast radius analysis for GCP — cross-project resource dependency mapping, org policy cascade effects, Shared VPC peering impact, Service Account impersonation chain analysis, and safe change sequencing.",
+  "source_type": "original",
+  "official_docs": [
+    "https://cloud.google.com/asset-inventory/docs/overview",
+    "https://cloud.google.com/vpc/docs/shared-vpc",
+    "https://cloud.google.com/iam/docs/understanding-service-accounts",
+    "https://cloud.google.com/resource-manager/docs/organization-policy/overview",
+    "https://cloud.google.com/vpc/docs/vpc-peering"
+  ],
+  "security_notes": "Cloud Asset Inventory requires roles/cloudasset.viewer — ensure the reviewing principal has this before attempting dependency analysis. Org policy changes with deny-override can lock out even org admins from specific resources — test in a non-production folder first.",
+  "last_verified": "2026-05-09",
+  "path": "skills/gcp/gcp-change-impact-advisor",
+  "author": "github: Raishin",
+  "version": "0.1.0"
+}

package/skills/gcp/gcp-change-impact-advisor/references/official-sources.md

@@ -0,0 +1,18 @@
+# Official sources
+
+Use this reference only when you need source grounding for GCP change impact and resource dependency service behavior or the detailed source list.
+
+## Google Cloud documentation
+
+Use these as starting points, not as proof of the user's live GCP state:
+- https://cloud.google.com/asset-inventory/docs/overview
+- https://cloud.google.com/vpc/docs/shared-vpc
+- https://cloud.google.com/iam/docs/understanding-service-accounts
+- https://cloud.google.com/resource-manager/docs/organization-policy/overview
+- https://cloud.google.com/vpc/docs/vpc-peering
+- https://cloud.google.com/asset-inventory/docs/searching-iam-policies
+- https://cloud.google.com/resource-manager/docs/creating-managing-folders
+
+## Grounding rule
+
+Official documentation explains GCP resource dependency and org policy behavior. It does not prove the user's current Shared VPC topology, SA impersonation chains, or org hierarchy. Prefer live GCP CLI/API evidence or sanitized user-provided evidence for current-state claims.

package/skills/gcp/gcp-change-impact-advisor/references/workflow-and-output.md

@@ -0,0 +1,62 @@
+# Workflow and output contract
+
+Use this reference only when performing the full change impact analysis, dependency mapping, or safe sequencing review.
+
+## Analysis domains
+
+Check these areas before giving a verdict:
+- Org policy cascade: constraint type (list vs. boolean), affected folder/project hierarchy, deny-override risk
+- Shared VPC impact: host project identification, enumeration of all service projects, subnet/firewall/route change scope
+- Service Account dependency chain: all IAM policy bindings, workload identity bindings, impersonation chains
+- Cloud Asset Inventory coverage: confirming `roles/cloudasset.viewer` is available before dependency analysis
+- VPC peering topology: direct peers only (non-transitive), route advertisement changes, firewall rule propagation
+- Change sequencing: dependencies between steps, blast radius reduction through staged rollout
+- Rollback design: reversibility of each change step, approval gates before irreversible actions
+
+## Safe workflow
+
+1. **Frame scope**
+   - Change type and target resource:
+   - Environment (prod/non-prod):
+   - Business criticality and owner:
+   - Required outcome:
+   - Explicit non-goals:
+2. **Collect evidence**
+   - Prefer live GCP CLI/API read-only evidence if available.
+   - Otherwise inspect repository IaC/config, sanitized user evidence, or official Google Cloud docs.
+   - Label each finding as `live evidence`, `repo evidence`, `user-provided evidence`, `documentation-based`, or `inference`.
+3. **Stress-test blast radius**
+   - What org folders and projects are in scope for org policy cascade?
+   - What Shared VPC service projects are attached to the affected host project?
+   - What workloads bind to or impersonate the affected Service Account?
+   - What VPC peers are directly affected (not transitively)?
+   - What evidence is missing?
+4. **Recommend the smallest safe action**
+   - Prefer narrow scope, staged rollout, validation, and rollback.
+   - Require explicit approval before irreversible changes (SA deletion, org policy enforcement mode).
+   - If the safest action is to stop and gather evidence, say that plainly.
+
+## Output contract
+
+Return this structure:
+```markdown
+# GCP Change Impact Advisor: <change description>
+## Executive verdict
+- Status: SAFE TO PROCEED / PROCEED WITH GATES / HIGH RISK / NEEDS EVIDENCE
+- Blast radius: <summary of affected resources and projects>
+- Biggest risk:
+- Evidence level:
+## Scope and assumptions
+- Confirmed:
+- Unknown:
+- Out of scope:
+## Findings
+| Severity | Finding | Evidence | Why it matters | Minimum safe action |
+|---|---|---|---|---|
+## Safe change sequencing
+1. <step> — approval gate: <gate>, rollback: <rollback>
+## Rollback plan
+- <step rollback or explicit irreversible note>
+## Residual risk
+- <risk or explicit none>
+```

package/skills/gcp/gcp-cloud-auth-advisor/SKILL.md

@@ -0,0 +1,69 @@
+---
+name: gcp-cloud-auth-advisor
+description: "Advise on Google Cloud authentication and authorization patterns — covering Application Default Credentials (ADC), service account best practices, Workload Identity Federation (for GKE pods and external workloads), human user auth (gcloud, IAP, Identity Platform), service-to-service auth (OIDC ID tokens, short-lived credentials), and anti-patterns like service account key downloads. Use when designing auth flows, debugging GCP auth failures, implementing least-privilege SA setup, or migrating from SA keys to keyless authentication."
+allowed-tools: Read Grep Glob
+metadata:
+  author: "github: Raishin"
+  version: "0.1.0"
+  updated: "2026-05-09"
+  category: security
+---
+
+# GCP Cloud Auth Advisor
+
+## Core Directive: Clarify Before Prescribing
+
+Ask 4 questions before providing a solution:
+1. Who/what is authenticating? (Human developer, local script, production workload, external cloud)
+2. Where is the code running? (Laptop, Compute Engine, GKE, Cloud Run, AWS/Azure/on-prem)
+3. What is the target? (Google Cloud API, custom app built on GCP)
+4. Are you using a high-level client library? (Python, Go, Node.js — usually handle ADC automatically)
+
+## Human Authentication Patterns
+
+- **Google-Managed Accounts** (Cloud Identity / Google Workspace) — managed lifecycle
+- **Federation** (GCDS sync with Active Directory / Entra ID)
+- **Workforce Identity Federation** — syncless, attribute-based SSO — recommended for enterprise
+- **Developer local access**: `gcloud auth login` (CLI auth), `gcloud auth application-default login` (ADC for client libraries)
+- **Service Account Impersonation**: use `--impersonate-service-account` instead of downloading SA keys for local dev
+- **End-user apps**: IAP for protecting internal apps without VPN; Identity Platform for consumer sign-in
+
+## Service-to-Service Authentication (Production)
+
+- Attach service account to compute resource (Compute Engine, Cloud Run, GKE) — access token provided via metadata server
+- NEVER use Service Account Keys in production — they are long-lived, hard to rotate, and a common breach vector
+- **GKE**: Workload Identity Federation for GKE — maps Kubernetes SA to Google SA; eliminates node-level SA key sharing
+- **External workloads** (AWS, Azure, on-prem): Workload Identity Federation — exchange external token for short-lived Google token; no keys needed
+- **Service-to-custom-app**: OIDC ID Token in `Authorization: Bearer` header — use `google.auth.transport.requests.AuthorizedSession` or equivalent
+
+## ADC Search Order
+
+`GOOGLE_APPLICATION_CREDENTIALS` env var → local gcloud ADC JSON → attached SA metadata server
+
+## Anti-Patterns (Flag Immediately If Seen)
+
+- SA keys downloaded and stored in code/environment → redirect to impersonation or WIF
+- Default Compute Engine SA used for production → create custom minimal-privilege SA
+- `0.0.0.0/0` authorized networks → restrict to known CIDRs
+- API keys with no restrictions → add API + application restrictions
+- Access scopes restricting token on GKE node pool → check SA IAM, not just scopes
+
+## Validation Checklist (Always Output at the End)
+
+- [ ] Local development: use gcloud ADC or SA impersonation, NOT SA keys
+- [ ] Production on GCP: attached SA, NOT key files
+- [ ] GKE: Workload Identity enabled, NOT node SA
+- [ ] External (AWS/Azure/on-prem): Workload Identity Federation, NOT cross-cloud SA keys
+- [ ] Custom app calls: OIDC ID tokens, NOT access tokens
+- [ ] API Keys: restricted to specific API + application
+
+## Official Docs
+
+- https://cloud.google.com/docs/authentication
+- https://cloud.google.com/iam/docs/workload-identity-federation
+- https://cloud.google.com/kubernetes-engine/docs/how-to/workload-identity
+- https://cloud.google.com/docs/authentication/application-default-credentials
+
+## Security Notes
+
+Read-only advisory. Never generate, store, or echo credentials, tokens, or service account keys. If a user pastes a key, flag it immediately as a security risk and advise rotation. Validate all auth designs against least-privilege principle.

package/skills/gcp/gcp-cloud-auth-advisor/metadata.json

@@ -0,0 +1,20 @@
+{
+  "id": "gcp-cloud-auth-advisor",
+  "name": "GCP Cloud Auth Advisor",
+  "type": "skill",
+  "provider": "gcp",
+  "harnesses": ["codex", "claude-code", "cursor", "gemini", "kiro", "other"],
+  "summary": "Advise on Google Cloud authentication and authorization patterns — covering ADC, service account best practices, Workload Identity Federation, human user auth, service-to-service auth, and anti-patterns like service account key downloads.",
+  "source_type": "original",
+  "official_docs": [
+    "https://cloud.google.com/docs/authentication",
+    "https://cloud.google.com/iam/docs/workload-identity-federation",
+    "https://cloud.google.com/kubernetes-engine/docs/how-to/workload-identity",
+    "https://cloud.google.com/docs/authentication/application-default-credentials"
+  ],
+  "security_notes": "Read-only advisory. Never generate, store, or echo credentials, tokens, or service account keys. If a user pastes a key, flag it immediately as a security risk and advise rotation. Validate all auth designs against least-privilege principle.",
+  "last_verified": "2026-05-09",
+  "path": "skills/gcp/gcp-cloud-auth-advisor",
+  "author": "github: Raishin",
+  "version": "0.1.0"
+}

package/skills/gcp/gcp-cloud-run-functions-operator/SKILL.md

@@ -0,0 +1,86 @@
+---
+name: gcp-cloud-run-functions-operator
+description: Deploy and operate Cloud Run services, Cloud Functions gen2, Eventarc triggers, traffic splitting for progressive delivery, and cold-start optimization strategies.
+allowed-tools: Read Grep Glob
+metadata:
+  author: "github: Raishin"
+  version: "0.2.0"
+  updated: "2026-05-09"
+  category: platform
+---
+
+# GCP Cloud Run and Functions Operator
+
+## Purpose
+
+Act as a rigorous Cloud Run and Cloud Functions operator. Keep serverless services reliable, cost-efficient, and free of cold-start surprises or silent VPC connectivity gaps.
+
+## Cloud Run Resource Types
+
+Cloud Run has three distinct resource types — confirm which one the user needs before proceeding:
+
+| Resource Type | Use When | Key Characteristics |
+|---|---|---|
+| **Service** | HTTP/gRPC API, event response, web app | Stateless, scale-to-zero, unique HTTPS endpoint, request-based billing |
+| **Job** | Batch processing, scheduled task, data pipeline step | Runs to completion, parallelizable tasks, no persistent endpoint |
+| **Worker Pool** | Pull-based consumers (Kafka, Pub/Sub pull, RabbitMQ) | Always-on, no HTTP endpoint, pulls work from queues |
+
+## Reference Directory
+
+| Scenario | Trigger Keywords | Reference |
+|---|---|---|
+| Deploy a service | HTTP, web app, API, deploy container, autoscale | [Services section](#services) |
+| Run a job | batch, scheduled, cron, run to completion, data pipeline | [Jobs section](#jobs) |
+| Worker pool setup | Kafka consumer, Pub/Sub pull, RabbitMQ, background worker | [Worker Pools section](#worker-pools) |
+| IAM & auth | invoke, service account, ingress, unauthenticated | [Security section](#security) |
+| VPC connectivity | VPC connector, egress, private IP, Cloud SQL, Memorystore | [Networking section](#networking) |
+| Cost & scaling | concurrency, min-instances, max-instances, cold start | [Scaling & Cost section](#scaling--cost) |
+
+## When to use
+
+Use this skill for:
+
+- Cloud Run service deployment, revision management, and traffic splitting
+- Cloud Run jobs for batch and scheduled workloads
+- Cloud Run worker pools for pull-based queue consumers (Kafka, Pub/Sub pull, RabbitMQ)
+- Cloud Functions gen2 deployment and configuration
+- Eventarc trigger design (Pub/Sub, GCS, Firestore, Audit Logs, custom sources)
+- Progressive delivery via revision traffic splits (canary, blue/green)
+- Cold-start analysis and minimum instances recommendations
+- Concurrency tuning and CPU allocation mode (request-only vs. always-on)
+- VPC connectivity (Direct VPC Egress vs. VPC connector) for private resource access
+
+## Key Cloud Run and Functions specifics
+
+- Cloud Run revision traffic: you can split traffic across multiple revisions (e.g., 90/10 canary) — this is the primary progressive delivery mechanism.
+- Minimum instances: prevents cold starts but costs money even when idle. Use for latency-sensitive services.
+- Cloud Functions gen2 runs on Cloud Run internally — same container model, same networking.
+- Eventarc: event-driven triggers from Pub/Sub, GCS, Firestore, Audit Logs, and custom sources. Use instead of polling patterns.
+- Concurrency: Cloud Run supports up to 1000 concurrent requests per instance. CPU is only allocated during request processing by default (not during idle).
+- Always-on CPU: required for background tasks or WebSockets — set cpu="always" to keep CPU allocated between requests.
+- VPC connector or Direct VPC Egress: required for Cloud Run to access private resources in a VPC. Direct VPC Egress is newer and preferred.
+
+## Lean operating rules
+
+- Prefer official GCP documentation and live evidence over memory or inference.
+- Separate confirmed facts from inference. If state was not queried or shown, say so.
+- Challenge missing min-instances for latency-sensitive services, CPU-only-on-request for background workloads, and missing VPC egress for private access.
+- Keep the answer scoped, reversible, least-privilege, and explicit about blockers or unknowns.
+- Load references only when needed; do not pull all deep guidance into short answers.
+
+## References
+
+Load these only when needed:
+
+- [Workflow and output contract](references/workflow-and-output.md) — use when executing the full review or formatting the final answer.
+- [Official sources](references/official-sources.md) — use when grounding Cloud Run/Functions behavior or checking the detailed source list.
+
+## Response minimum
+
+Return, at minimum:
+
+- the scoped target and evidence level,
+- the main risks or control gaps,
+- the safest next actions,
+- validation or rollback notes where relevant,
+- the assumptions or blockers that prevent stronger conclusions.

package/skills/gcp/gcp-cloud-run-functions-operator/metadata.json

@@ -0,0 +1,27 @@
+{
+  "id": "gcp-cloud-run-functions-operator",
+  "name": "GCP Cloud Run and Functions Operator",
+  "type": "skill",
+  "provider": "gcp",
+  "harnesses": [
+    "codex",
+    "claude-code",
+    "cursor",
+    "gemini",
+    "kiro",
+    "other"
+  ],
+  "summary": "Deploy and operate Cloud Run services, Cloud Functions gen2, Eventarc triggers, traffic splitting for progressive delivery, and cold-start optimization strategies.",
+  "source_type": "original",
+  "official_docs": [
+    "https://cloud.google.com/run/docs/overview/what-is-cloud-run",
+    "https://cloud.google.com/run/docs/rollouts-rollbacks-traffic-migration",
+    "https://cloud.google.com/functions/docs/concepts/overview",
+    "https://cloud.google.com/eventarc/docs/overview"
+  ],
+  "security_notes": "Always-on CPU is required for background tasks or WebSockets. Direct VPC Egress is preferred over VPC connector for Cloud Run private VPC access. CPU is not allocated during idle by default.",
+  "last_verified": "2026-05-08",
+  "path": "skills/gcp/gcp-cloud-run-functions-operator",
+  "author": "github: Raishin",
+  "version": "0.2.0"
+}

package/skills/gcp/gcp-cloud-run-functions-operator/references/official-sources.md

@@ -0,0 +1,15 @@
+# Official sources
+
+Use this reference only when you need source grounding for Cloud Run or Cloud Functions behavior or the detailed source list.
+
+## GCP documentation
+
+Use these as starting points, not as proof of the user's live GCP state:
+- https://cloud.google.com/run/docs/overview/what-is-cloud-run
+- https://cloud.google.com/run/docs/rollouts-rollbacks-traffic-migration
+- https://cloud.google.com/functions/docs/concepts/overview
+- https://cloud.google.com/eventarc/docs/overview
+
+## Grounding rule
+
+Official documentation explains Cloud Run and Cloud Functions behavior. It does not prove the user's current revision traffic splits, minimum instance configuration, concurrency settings, or VPC egress setup. Prefer live GCP CLI/API evidence or sanitized user-provided evidence for current-state claims.

package/skills/gcp/gcp-cloud-run-functions-operator/references/workflow-and-output.md

@@ -0,0 +1,61 @@
+# Workflow and output contract
+
+Use this reference only when performing the full review, implementation guidance, or production-readiness pass.
+
+## Review domains
+
+Check these areas before giving a verdict:
+- Service/function inventory (region, revision count, current traffic split)
+- Revision health (container startup time, error rate, latency p99)
+- Cold-start profile (min-instances setting vs. latency SLA)
+- Concurrency settings (max-instances, max-concurrency, CPU allocation mode)
+- VPC connectivity (Direct VPC Egress vs. VPC connector vs. no VPC)
+- Eventarc triggers (source, filter, delivery guarantees)
+- IAM (invoker bindings, service account least privilege)
+
+## Safe workflow
+
+1. **Frame scope**
+   - Project/region and service or function name:
+   - Traffic split and active revisions:
+   - Latency SLA and cold-start tolerance:
+   - Required outcome:
+   - Explicit non-goals:
+2. **Collect evidence**
+   - Prefer live GCP CLI/API read-only evidence if available.
+   - Otherwise inspect repository IaC/config, sanitized user evidence, or official GCP docs.
+   - Label each finding as `live evidence`, `repo evidence`, `user-provided evidence`, `documentation-based`, or `inference`.
+3. **Stress-test risk**
+   - Are latency-sensitive services running without minimum instances?
+   - Are background services missing always-on CPU?
+   - Are services that need private VPC access missing Direct VPC Egress?
+   - Are all invoker bindings scoped to the minimum necessary identities?
+   - What evidence is missing?
+4. **Recommend the smallest safe action**
+   - Prefer narrow scope, staged rollout, validation, and rollback.
+   - If the safest action is to stop and gather evidence, say that plainly.
+
+## Output contract
+
+Return this structure:
+```markdown
+# GCP Cloud Run and Functions Operator: <scope>
+## Executive verdict
+- Status: READY / READY WITH RISKS / NOT READY / NEEDS EVIDENCE
+- Biggest risk:
+- Evidence level:
+## Scope and assumptions
+- Confirmed:
+- Unknown:
+- Out of scope:
+## Findings
+| Severity | Finding | Evidence | Why it matters | Minimum safe action |
+|---|---|---|---|---|
+## Recommended actions
+1. <action> — owner: <owner>, validation: <check>, rollback: <rollback>
+## Validation
+- Commands or checks:
+- Expected result:
+## Residual risk
+- <risk or explicit none>
+```

package/skills/gcp/gcp-cloudbuild-deploy-cicd-operator/SKILL.md

@@ -0,0 +1,58 @@
+---
+name: gcp-cloudbuild-deploy-cicd-operator
+description: Build and operate CI/CD pipelines using Cloud Build, Cloud Deploy delivery pipelines, Artifact Registry, SLSA provenance generation, and release gating with approval workflows.
+allowed-tools: Read Grep Glob
+metadata:
+  author: "github: Raishin"
+  version: "0.1.0"
+  updated: "2026-05-08"
+  category: delivery
+---
+
+# GCP Cloud Build Deploy CI/CD Operator
+
+## Purpose
+
+Act as the GCP CI/CD operator who enforces supply chain security, least-privilege build accounts, and progressive delivery gating.
+
+## When to use
+
+Use this skill for:
+
+- Cloud Build pipeline design, cloudbuild.yaml authoring, private pool VPC configuration
+- Cloud Deploy delivery pipeline setup for GKE, Cloud Run, and Anthos (dev→staging→prod progression)
+- Artifact Registry repository management (Docker, Maven, npm, Python, Helm) and retention policy configuration
+- SLSA provenance generation and Binary Authorization enforcement
+- Cloud Build service account permission audits (over-privilege is a common security gap)
+- Skaffold version compatibility management with Cloud Deploy
+- Approval workflow and release gate configuration
+
+## Lean operating rules
+
+- Prefer live GCP evidence from sanitized gcloud / Cloud Build API output when available; otherwise use official Google Cloud documentation.
+- Cloud Build service account minimum permissions: Cloud Run Admin + Artifact Registry Writer + GKE Developer. Over-privileged build service accounts are a common supply chain security gap.
+- SLSA provenance combined with Binary Authorization is required for supply chain enforcement — provenance alone is insufficient.
+- Artifact Registry is preferred over Container Registry (GCR) — confirm which registry is in use before making retention or region recommendations.
+- Artifact Registry is regional, not global — latency to the build region matters for large images.
+- Skaffold version must be compatible with the Cloud Deploy release version — mismatch causes silent rendering failures.
+- Separate confirmed facts from inference. If state was not queried or shown, say so.
+- Challenge broad IAM roles, public exposure, destructive automation, untested recovery, hidden cost, and vague production claims.
+- Keep the answer scoped, reversible, least-privilege, and explicit about blockers or unknowns.
+- Load references only when needed; do not pull all deep guidance into short answers.
+
+## References
+
+Load these only when needed:
+
+- [Workflow and output contract](references/workflow-and-output.md) — use when executing the full review, pipeline audit, implementation guidance, or formatting the final answer.
+- [Official sources](references/official-sources.md) — use when grounding GCP CI/CD service behavior or checking the detailed source list.
+
+## Response minimum
+
+Return, at minimum:
+
+- the scoped target and evidence level,
+- the main risks or control gaps (especially service account permissions and supply chain),
+- the safest next actions,
+- validation or rollback notes where relevant,
+- the assumptions or blockers that prevent stronger conclusions.

package/skills/gcp/gcp-cloudbuild-deploy-cicd-operator/metadata.json

@@ -0,0 +1,27 @@
+{
+  "id": "gcp-cloudbuild-deploy-cicd-operator",
+  "name": "GCP Cloud Build Deploy CI/CD Operator",
+  "type": "skill",
+  "provider": "gcp",
+  "harnesses": [
+    "codex",
+    "claude-code",
+    "cursor",
+    "gemini",
+    "kiro",
+    "other"
+  ],
+  "summary": "Build and operate CI/CD pipelines using Cloud Build, Cloud Deploy delivery pipelines, Artifact Registry, SLSA provenance generation, and release gating with approval workflows.",
+  "source_type": "original",
+  "official_docs": [
+    "https://cloud.google.com/build/docs/overview",
+    "https://cloud.google.com/deploy/docs/overview",
+    "https://cloud.google.com/artifact-registry/docs/overview",
+    "https://cloud.google.com/build/docs/securing-builds/view-build-provenance"
+  ],
+  "security_notes": "Cloud Build service accounts are commonly over-privileged — minimum required permissions are Cloud Run Admin + Artifact Registry Writer + GKE Developer. SLSA provenance combined with Binary Authorization prevents tampered artifacts from reaching production.",
+  "last_verified": "2026-05-08",
+  "path": "skills/gcp/gcp-cloudbuild-deploy-cicd-operator",
+  "author": "github: Raishin",
+  "version": "0.1.0"
+}

package/skills/gcp/gcp-cloudbuild-deploy-cicd-operator/references/official-sources.md

@@ -0,0 +1,19 @@
+# Official sources
+
+Use this reference only when you need source grounding for GCP CI/CD service behavior or the detailed source list.
+
+## Google Cloud documentation
+
+Use these as starting points, not as proof of the user's live GCP state:
+- https://cloud.google.com/build/docs/overview
+- https://cloud.google.com/deploy/docs/overview
+- https://cloud.google.com/artifact-registry/docs/overview
+- https://cloud.google.com/build/docs/securing-builds/view-build-provenance
+- https://cloud.google.com/binary-authorization/docs/overview
+- https://cloud.google.com/deploy/docs/config-files
+- https://cloud.google.com/build/docs/configuring-builds/create-basic-configuration
+- https://cloud.google.com/deploy/docs/approve-rollout
+
+## Grounding rule
+
+Official documentation explains GCP Cloud Build and Cloud Deploy service behavior. It does not prove the user's current project, pipeline state, service account permissions, artifact configurations, or operational posture. Prefer live GCP CLI/API evidence or sanitized user-provided evidence for current-state claims.

package/skills/gcp/gcp-cloudbuild-deploy-cicd-operator/references/workflow-and-output.md

@@ -0,0 +1,60 @@
+# Workflow and output contract
+
+Use this reference only when performing the full CI/CD pipeline review, security audit, implementation guidance, or production-readiness pass.
+
+## Review domains
+
+Check these areas before giving a verdict:
+- Cloud Build: cloudbuild.yaml structure, private pool vs. public pool, build trigger inventory
+- Cloud Deploy: delivery pipeline stages, target configurations, promotion criteria
+- Artifact Registry: repository type, region, retention policy, public access settings
+- Service accounts: Cloud Build SA permissions (minimum: Cloud Run Admin + Artifact Registry Writer + GKE Developer)
+- SLSA provenance: provenance generation enabled, Binary Authorization policy configured
+- Approval gates: required approvers per stage, timeout configuration
+- Skaffold: version pinned and compatible with Cloud Deploy release version
+
+## Safe workflow
+
+1. **Frame scope**
+   - Project/region/environment:
+   - Business criticality and owner:
+   - Data classification and compliance driver:
+   - Required outcome:
+   - Explicit non-goals:
+2. **Collect evidence**
+   - Prefer live GCP CLI/API read-only evidence if available.
+   - Otherwise inspect repository IaC/config (cloudbuild.yaml, skaffold.yaml), sanitized user evidence, or official Google Cloud docs.
+   - Label each finding as `live evidence`, `repo evidence`, `user-provided evidence`, `documentation-based`, or `inference`.
+3. **Stress-test risk**
+   - What service account permissions exceed minimum required?
+   - What artifacts can be deployed without provenance verification?
+   - What stages lack approval gates?
+   - What evidence is missing?
+4. **Recommend the smallest safe action**
+   - Prefer narrow scope, staged rollout, validation, and rollback.
+   - If the safest action is to stop and gather evidence, say that plainly.
+
+## Output contract
+
+Return this structure:
+```markdown
+# GCP Cloud Build Deploy CI/CD Operator: <scope>
+## Executive verdict
+- Status: READY / READY WITH RISKS / NOT READY / NEEDS EVIDENCE
+- Biggest risk:
+- Evidence level:
+## Scope and assumptions
+- Confirmed:
+- Unknown:
+- Out of scope:
+## Findings
+| Severity | Finding | Evidence | Why it matters | Minimum safe action |
+|---|---|---|---|---|
+## Recommended actions
+1. <action> — owner: <owner>, validation: <check>, rollback: <rollback>
+## Validation
+- Commands or checks:
+- Expected result:
+## Residual risk
+- <risk or explicit none>
+```