菜单
Inspect and validate Vertex AI Agent Engine deployments including Code Execution Sandbox, Memory Bank, A2A protocol compliance, and security posture. Generates production readiness scores. Use when asked to "inspect agent engine" or "validate depl...
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
The model must invoke this skill when any trigger occurs - (1) user mentions "clang-format" or ".clang-format", (2) user requests analyzing code style/formatting patterns/conventions, (3) user requests creating/modifying/generating formatting configuration, (4) user troubleshoots formatting behavior or unexpected results, (5) user asks about brace styles/indentation/spacing/alignment/line breaking/pointer alignment, (6) user wants to preserve existing style/minimize whitespace changes/reduce formatting diffs/codify dominant conventions.
When setting up commit message validation for a project. When project has commitlint.config.js or .commitlintrc files. When configuring CI/CD to enforce commit format. When extracting commit rules for LLM prompt generation. When debugging commit message rejection errors.
The model must apply when tasks involve .gitlab-ci.yml configuration, GitLab Flavored Markdown (GLFM) syntax, gitlab-ci-local testing, CI/CD pipeline optimization, GitLab CI Steps composition, Docker-in-Docker workflows, or GitLab documentation creation. Triggers include modifying pipelines, writing GitLab README/Wiki content, debugging CI jobs locally, implementing caching strategies, or configuring release workflows.
This skill provides comprehensive documentation for Hatchling, the modern Python build backend that implements PEP 517/518/621/660 standards. Use this skill when working with Hatchling configuration, build system setup, Python packaging, pyproject.toml configuration, project metadata, dependencies, entry points, build hooks, version management, wheel and sdist builds, package distribution, setuptools migration, and troubleshooting Hatchling build errors.
When setting up local LLM inference without cloud APIs. When running GGUF models locally. When needing OpenAI-compatible API from a local model. When building offline/air-gapped AI tools. When troubleshooting local LLM server connections.
When setting up automated code quality checks on git commit. When project has .pre-commit-config.yaml. When implementing git hooks for formatting, linting, or validation. When creating prepare-commit-msg hooks to modify commit messages. When distributing a tool as a pre-commit hook.
| name | vertex-engine-inspector |
| description | Inspect and validate Vertex AI Agent Engine deployments including Code Execution Sandbox, Memory Bank, A2A protocol compliance, and security posture. Generates production readiness scores. Use when asked to "inspect agent engine" or "validate depl... |
| allowed-tools | Read, Grep, Glob, Bash |
| license | MIT |
Expert inspector for the Vertex AI Agent Engine managed runtime. Performs comprehensive validation of deployed agents including runtime configuration, security posture, performance settings, A2A protocol compliance, and production readiness scoring.
Critical Checks:
✅ State TTL between 7-14 days (optimal for production)
✅ Sandbox type is SECURE_ISOLATED
✅ IAM permissions limited to required GCP services only
✅ Timeout configured appropriately
⚠️ State TTL < 7 days may cause premature session loss
❌ State TTL > 14 days not allowed by Agent Engine
Critical Checks:
✅ Max memories >= 100 (prevents conversation truncation)
✅ Indexing enabled (fast query performance)
✅ Auto-cleanup enabled (prevents quota exhaustion)
✅ Encrypted at rest (Firestore default)
⚠️ Low memory limit may truncate long conversations
/.well-known/agent-cardPOST /v1/tasks:send responds correctlyGET /v1/tasks/{task_id} accessibleCompliance Report:
✅ AgentCard accessible and valid
✅ Task submission API functional
✅ Status polling API functional
✅ Protocol version 1.0
❌ Missing AgentCard fields: [...]
❌ Task API not responding (check IAM/networking)
Security Score:
🟢 SECURE (90-100%): Production ready
🟡 NEEDS ATTENTION (70-89%): Address issues before prod
🔴 INSECURE (<70%): Do not deploy to production
Health Status:
🟢 HEALTHY: Error rate < 5%, latency < 3s (p95)
🟡 DEGRADED: Error rate 5-10% or latency 3-5s
🔴 UNHEALTHY: Error rate > 10% or latency > 5s
Observability Score:
✅ All 5 pillars configured: Metrics, Logs, Traces, Alerts, Dashboards
⚠️ Missing alerts for critical scenarios
❌ No monitoring configured (production blocker)
| Category | Weight | Checks |
|---|---|---|
| Security | 30% | 6 checks (IAM, VPC-SC, encryption, etc.) |
| Performance | 25% | 6 checks (scaling, limits, SLOs, etc.) |
| Monitoring | 20% | 6 checks (dashboards, alerts, logs, etc.) |
| Compliance | 15% | 5 checks (audit logs, DR, privacy, etc.) |
| Reliability | 10% | 5 checks (multi-region, failover, etc.) |
🟢 PRODUCTION READY (85-100%)
- All critical checks passed
- Minor optimizations recommended
- Safe to deploy
🟡 NEEDS IMPROVEMENT (70-84%)
- Some important checks failed
- Address issues before production
- Staging deployment acceptable
🔴 NOT READY (<70%)
- Critical failures present
- Do not deploy to production
- Fix blocking issues first
1. Connect to Agent Engine
2. Retrieve agent metadata
3. Parse runtime configuration
4. Extract Code Execution settings
5. Extract Memory Bank settings
6. Document VPC configuration
1. Test AgentCard endpoint
2. Validate AgentCard structure
3. Test Task API (POST /v1/tasks:send)
4. Test Status API (GET /v1/tasks/{id})
5. Verify A2A protocol version
1. Review IAM roles and permissions
2. Check VPC Service Controls
3. Validate encryption settings
4. Scan for hardcoded secrets
5. Verify Model Armor enabled
6. Assess service account security
1. Query Cloud Monitoring metrics
2. Calculate error rate (last 24h)
3. Analyze latency percentiles
4. Review token usage and costs
5. Check auto-scaling behavior
6. Validate resource limits
1. Run all checklist items (28 checks)
2. Calculate category scores
3. Calculate overall score
4. Determine readiness status
5. Generate recommendations
6. Create action plan
Read-only inspection - Cannot modify configurations:
Agent ID: gcp-deployer-agent
Deployment Status: RUNNING
Inspection Date: 2025-12-09
Runtime Configuration:
Model: gemini-2.5-flash
Code Execution: ✅ Enabled (TTL: 14 days)
Memory Bank: ✅ Enabled (retention: 90 days)
VPC: ✅ Configured (private-vpc-prod)
A2A Protocol Compliance:
AgentCard: ✅ Valid
Task API: ✅ Functional
Status API: ✅ Functional
Protocol Version: 1.0
Security Posture:
IAM: ✅ Least privilege (score: 95%)
VPC-SC: ✅ Enabled
Model Armor: ✅ Enabled
Encryption: ✅ At-rest & in-transit
Overall: 🟢 SECURE (92%)
Performance Metrics (24h):
Request Count: 12,450
Error Rate: 2.3% 🟢
Latency (p95): 1,850ms 🟢
Token Usage: 450K tokens
Cost Estimate: $12.50/day
Production Readiness:
Security: 92% (28/30 points)
Performance: 88% (22/25 points)
Monitoring: 95% (19/20 points)
Compliance: 80% (12/15 points)
Reliability: 70% (7/10 points)
Overall Score: 87% 🟢 PRODUCTION READY
Recommendations:
1. Enable multi-region deployment (reliability +10%)
2. Configure automated backups (compliance +5%)
3. Add circuit breaker pattern (reliability +5%)
4. Optimize memory bank indexing (performance +3%)
Inspector checks:
Inspector checks:
Inspector checks: