agent-skill-kit 3.9.135

package/.agent/skills/observability/SKILL.md

@@ -0,0 +1,178 @@
+---
+name: observability
+description: >-
+  OpenTelemetry-based observability: unified logs, metrics, and distributed traces.
+  Use when setting up monitoring, adding telemetry, or configuring observability stacks.
+  NOT for server management (use server-ops) or performance optimization (use perf-optimizer).
+category: devops-architect
+triggers: ["monitoring", "observability", "OpenTelemetry", "telemetry", "instrumentation"]
+coordinates_with: ["server-ops", "cicd-pipeline", "problem-checker"]
+success_metrics: ["0 Telemetry Errors", "100% Correlation"]
+metadata:
+  author: pikakit
+  version: "3.9.134"
+---
+
+# Observability — Unified Telemetry with OpenTelemetry
+
+> 3 pillars. Vendor-agnostic. Fixed sampling. Auto-instrumentation first.
+
+**Key:** Set up once, benefit everywhere (logs, metrics, traces share same SDK).
+
+---
+
+## 5 Must-Ask Questions (Before Any Setup)
+
+| # | Question | Options |
+|---|----------|---------|
+| 1 | Runtime Environment? | Node.js, Python, Go |
+| 2 | Telemetry Provider? | Datadog, Grafana, Sentry, OSS |
+| 3 | Desired Coverage? | Logs, Metrics, Traces or All |
+| 4 | Deployment Target? | K8s, Serverless, VM |
+| 5 | Privacy Constraints? | PII masking, GDPR, HIPAA |
+
+---
+
+## Prerequisites
+
+| Runtime | Install Command |
+|---------|----------------|
+| Node.js | `npm install @opentelemetry/sdk-node @opentelemetry/auto-instrumentations-node` |
+| Python | `pip install opentelemetry-sdk opentelemetry-instrumentation` |
+
+---
+
+## When to Use
+
+| Situation | Action |
+|-----------|--------|
+| Need unified telemetry | Set up OpenTelemetry SDK |
+| Debug production issues | Check traces (trace_id) |
+| Monitor performance | Use metrics |
+| Track errors | Configure structured logging |
+
+---
+
+## System Boundaries
+
+| Owned by This Skill | NOT Owned |
+|---------------------|-----------|
+| SDK configuration (Node.js, Python) | Infrastructure provisioning (→ server-ops) |
+| Sampling strategy (3 environments) | Alerting rules (→ /monitor workflow) |
+| Provider routing (3 providers) | CI/CD monitoring (→ cicd-pipeline) |
+| Auto-instrumentation guidance | Dashboard design |
+
+**Expert decision skill:** Produces configuration guidance. Does not install packages.
+
+---
+
+## Three Pillars (Unified via trace_id)
+
+| Pillar | Purpose | Example |
+|--------|---------|---------|
+| **Logs** | Events with context | "User login failed" |
+| **Metrics** | Numerical measurements | "CPU usage: 45%" |
+| **Traces** | Request flow | "API call took 250ms" |
+
+---
+
+## Sampling by Environment (Fixed)
+
+| Environment | Rate | Rationale |
+|-------------|------|-----------|
+| Development | 100% | Full visibility for debugging |
+| Staging | 50% | Balance cost vs visibility |
+| Production | 1-10% | Cost control |
+
+---
+
+## Auto-Instrumented Libraries (Fixed)
+
+| Library | Data Captured |
+|---------|--------------|
+| HTTP | Request/response, status codes |
+| Express | Routes, middleware timing |
+| Prisma | Database queries |
+| Redis | Commands, latency |
+| Next.js | Page loads, API routes |
+
+---
+
+## Provider Integration (Fixed)
+
+| Provider | Exporter |
+|----------|----------|
+| Datadog | `@opentelemetry/exporter-datadog` |
+| Grafana | OTLP HTTP with auth |
+| Sentry | `@sentry/node` |
+
+---
+
+## Audit Logging (OpenTelemetry)
+
+| Event | Metadata Payload | Severity |
+|-------|------------------|----------|
+| `architecture_decision` | `{"runtime": "...", "provider": "..."}` | `INFO` |
+| `instrumentation_plan` | `{"auto_captured": 5, "manual": 2}` | `INFO` |
+| `build_verification` | `{"status": "pass|fail", "metrics_met": true}` | `INFO` |
+
+All executions MUST emit the `build_verification` span before reporting completion.
+
+---
+
+## Error Taxonomy
+
+| Code | Recoverable | Trigger |
+|------|-------------|---------|
+| `ERR_INVALID_REQUEST_TYPE` | No | Request type not supported |
+| `ERR_UNKNOWN_RUNTIME` | Yes | Runtime not nodejs or python |
+| `ERR_UNKNOWN_ENVIRONMENT` | Yes | Environment not recognized |
+| `ERR_UNKNOWN_PROVIDER` | Yes | Provider not one of 3 |
+
+**Zero internal retries.** Deterministic; same context = same configuration.
+
+---
+
+## Anti-Patterns
+
+| ❌ Don't | ✅ Do |
+|---------|-------|
+| Create excessive custom spans | Use auto-instrumentation first |
+| Put sensitive data in span tags | Sanitize all tags before export |
+| Hardcode SERVICE_NAME | Use environment variable |
+| Sample 100% in production | Use 1-10% for production |
+| Ignore errors in telemetry | Log all exceptions |
+| Ignore IDE warnings/errors | Call `problem-checker` to auto-fix |
+
+---
+
+## Troubleshooting
+
+| Problem | Solution |
+|---------|----------|
+| No traces appearing | Check `OTEL_EXPORTER_OTLP_ENDPOINT` is set |
+| High cardinality metrics | Reduce label values, use histograms |
+| Missing spans | Verify auto-instrumentation is loaded |
+| Prometheus scrape fails | Check `/metrics` endpoint and firewall |
+
+---
+
+## 📑 Content Map
+
+| File | Description | When to Read |
+|------|-------------|--------------|
+| [engineering-spec.md](rules/engineering-spec.md) | Full engineering spec | Architecture review |
+
+---
+
+## 🔗 Related
+
+| Item | Type | Purpose |
+|------|------|---------|
+| `/monitor` | Workflow | Monitoring setup |
+| `server-ops` | Skill | Server management |
+| `cicd-pipeline` | Skill | Deployment monitoring |
+
+---
+
+⚡ PikaKit v3.9.134

package/.agent/skills/observability/rules/_sections.md

@@ -0,0 +1,15 @@
+# Sections
+
+This file defines all sections, their ordering, impact levels, and descriptions.
+The section ID (in parentheses) is the filename prefix used to group rules.
+
+---
+
+## 1. Engineering Specification (engineering)
+
+**Impact:** MEDIUM  
+**Description:** Full engineering specification covering contracts, security, and scalability.
+
+---
+
+⚡ PikaKit v3.9.134

package/.agent/skills/observability/rules/_template.md

@@ -0,0 +1,32 @@
+---
+title: Rule Title Here
+impact: MEDIUM
+impactDescription: Optional description of impact (e.g., "20-50% improvement")
+tags: tag1, tag2
+---
+
+## Rule Title Here
+
+**Impact: MEDIUM (optional impact description)**
+
+Brief explanation of the rule and why it matters. This should be clear and concise, explaining the performance implications.
+
+**Incorrect (description of what's wrong):**
+
+```typescript
+// Bad code example here
+const bad = example()
+```
+
+**Correct (description of what's right):**
+
+```typescript
+// Good code example here
+const good = example()
+```
+
+Reference: [Link to documentation or resource](https://example.com)
+
+---
+
+⚡ PikaKit v3.9.134

package/.agent/skills/observability/rules/engineering-spec.md

@@ -0,0 +1,440 @@
+---
+title: Observability — Engineering Specification
+impact: MEDIUM
+tags: observability
+---
+
+# Observability — Engineering Specification
+
+> Production-grade specification for OpenTelemetry-based observability at FAANG scale.
+
+---
+
+## 1. Overview
+
+Observability provides structured decision frameworks for production telemetry: three pillars (logs, metrics, traces), OpenTelemetry SDK setup (Node.js, Python), auto-instrumentation routing, environment-based sampling, provider integration, and troubleshooting. The skill operates as an **Expert (decision tree)** — it produces SDK configuration decisions, sampling strategies, provider selection, and instrumentation guidance. It does not install packages, configure infrastructure, or deploy collectors.
+
+**Contract Version:** 2.0.0
+**Backward Compatibility:** breaking (first hardened version)
+**Breaking Changes:** None — new spec for first hardening
+
+---
+
+## 2. Problem Statement
+
+Production observability at scale faces four quantified problems:
+
+| Problem | Measurement | Impact |
+|---------|-------------|--------|
+| No telemetry correlation | 55% of apps lack trace-log-metric linking | Cannot trace request flow |
+| Vendor lock-in | 40% of apps use proprietary SDKs | Costly migration |
+| Trace oversampling in production | 30% of apps sample 100% in prod | Excessive cost |
+| Missing auto-instrumentation | 45% of apps rely on manual spans | Incomplete coverage |
+
+Observability eliminates these with OpenTelemetry (vendor-agnostic), unified correlation (trace_id across all pillars), fixed sampling rates by environment, and auto-instrumentation-first guidance.
+
+---
+
+## 3. Design Goals
+
+| ID | Goal | Measurable Constraint |
+|----|------|-----------------------|
+| G1 | Three pillars unified | Logs + Metrics + Traces via single SDK |
+| G2 | Vendor-agnostic | OpenTelemetry SDK; switchable exporters |
+| G3 | Fixed sampling | Dev: 100%, Staging: 50%, Prod: 1-10% |
+| G4 | Auto-instrumentation first | 5+ libraries auto-captured |
+| G5 | Provider routing | 3 providers: Datadog, Grafana, Sentry |
+| G6 | SDK quickstart | Node.js + Python setup in < 10 lines |
+
+---
+
+## 4. Non-Goals
+
+| ID | Excluded | Rationale |
+|----|----------|-----------|
+| NG1 | Infrastructure provisioning | Owned by `server-ops` skill |
+| NG2 | Alerting rules | Owned by `/monitor` workflow |
+| NG3 | Incident response | Owned by incident-response patterns |
+| NG4 | CI/CD monitoring | Owned by `cicd-pipeline` skill |
+| NG5 | Custom collector deployment | Infrastructure concern |
+| NG6 | Dashboard design | Provider-specific UI |
+
+---
+
+## 5. System Boundaries
+
+| Boundary | Owned | Not Owned |
+|----------|-------|-----------|
+| SDK configuration (Node.js, Python) | Setup guidance | Package installation |
+| Sampling strategy (3 environments) | Rate selection | Collector configuration |
+| Provider integration (3 providers) | Exporter selection | Provider account setup |
+| Auto-instrumentation routing | Library list | Library internals |
+| Three-pillar correlation | trace_id propagation guidance | Collector pipeline |
+
+**Side-effect boundary:** Observability produces SDK configuration guidance, sampling decisions, and provider recommendations. It does not install packages, modify infrastructure, or create dashboards.
+
+---
+
+## 6. Integration Model
+
+### 6.1 Agent Contract
+
+#### Input Schema
+
+```
+Request_Type: string          # "sdk-setup" | "sampling" | "provider" | "instrumentation" |
+                              # "troubleshoot" | "full-guide"
+Context: {
+  runtime: string             # "nodejs" | "python"
+  environment: string         # "development" | "staging" | "production"
+  provider: string | null     # "datadog" | "grafana" | "sentry" | null
+  libraries: Array<string> | null  # Libraries in use (e.g., ["express", "prisma"])
+  issue: string | null        # Troubleshooting issue description
+}
+contract_version: string      # "2.0.0"
+```
+
+#### Output Schema
+
+```
+Status: "success" | "error"
+Data: {
+  sdk_setup: {
+    runtime: string           # "nodejs" | "python"
+    install_command: string
+    config_snippet: string    # SDK initialization code
+    service_name_env: string  # "SERVICE_NAME"
+  } | null
+  sampling: {
+    environment: string
+    rate_percent: number      # 1-100
+    rationale: string
+  } | null
+  provider: {
+    name: string
+    exporter_package: string
+    config_notes: string
+  } | null
+  instrumentation: {
+    auto_captured: Array<{
+      library: string
+      data_captured: string
+    }>
+    manual_required: Array<string> | null
+  } | null
+  troubleshooting: {
+    problem: string
+    solution: string
+  } | null
+  metadata: {
+    contract_version: string
+    backward_compatibility: string
+  }
+}
+Error: ErrorSchema | null
+```
+
+#### Error Schema
+
+```
+Code: string                  # From Error Taxonomy (Section 11)
+Message: string
+Request_Type: string
+Recoverable: boolean
+```
+
+#### Deterministic Guarantees
+
+- Sampling is fixed: development → 100%, staging → 50%, production → 1-10%.
+- SDK setup is deterministic per runtime: Node.js → `@opentelemetry/sdk-node`, Python → `opentelemetry-sdk`.
+- Provider mapping is fixed: Datadog → `@opentelemetry/exporter-datadog`, Grafana → OTLP HTTP, Sentry → `@sentry/node`.
+- Auto-instrumentation libraries are fixed: HTTP, Express, Prisma, Redis, Next.js.
+- Three pillars always unified via trace_id.
+
+#### What Agents May Assume
+
+- OpenTelemetry is the standard SDK for all telemetry.
+- Auto-instrumentation captures HTTP, Express, Prisma, Redis, Next.js without manual code.
+- Sampling rates are fixed per environment.
+- trace_id correlates logs, metrics, and traces.
+
+#### What Agents Must NOT Assume
+
+- Packages are installed.
+- Collector infrastructure exists.
+- Provider accounts are configured.
+- 100% sampling is acceptable in production.
+- Custom spans are needed when auto-instrumentation covers the library.
+
+#### Side-Effect Boundaries
+
+| Operation | Side Effects |
+|-----------|-------------|
+| SDK setup | None; code snippet output |
+| Sampling | None; rate recommendation |
+| Provider | None; exporter recommendation |
+| Instrumentation | None; library list |
+| Troubleshoot | None; solution text |
+| Full guide | None; combined output |
+
+### 6.2 Workflow Contract
+
+#### Invocation Pattern
+
+```
+1. Identify runtime (Node.js or Python) and environment
+2. Invoke sdk-setup for configuration
+3. Invoke sampling for rate selection
+4. Invoke provider for exporter selection (if provider known)
+5. Invoke instrumentation for auto-captured library list
+6. Install and configure (caller's responsibility)
+```
+
+#### Execution Guarantees
+
+- Each invocation produces a complete recommendation.
+- All decisions are independent (can be invoked in any order).
+
+#### Failure Propagation Model
+
+| Failure Severity | Propagation | Workflow Action |
+|-----------------|-------------|-----------------|
+| Unknown runtime | Return error | Specify nodejs or python |
+| Unknown environment | Return error | Specify dev, staging, or production |
+| Unknown provider | Return error | Specify datadog, grafana, or sentry |
+
+#### Retry Boundaries
+
+- Zero internal retries. Deterministic output.
+
+#### Isolation Model
+
+- Each invocation is stateless and independent.
+
+#### Idempotency Expectations
+
+| Operation | Idempotent | Notes |
+|-----------|-----------|-------|
+| SDK setup | Yes | Same runtime = same config |
+| Sampling | Yes | Same environment = same rate |
+| Provider | Yes | Same provider = same exporter |
+| Instrumentation | Yes | Fixed library list |
+| Troubleshoot | Yes | Same issue = same solution |
+
+---
+
+## 7. Execution Model
+
+### 2-Phase Lifecycle
+
+| Phase | Action | Output |
+|-------|--------|--------|
+| **Classify** | Validate runtime, environment, request type | Classification |
+| **Recommend** | Generate SDK config, sampling rate, or provider guidance | Complete output |
+
+All phases synchronous. No async pipeline.
+
+---
+
+## 8. Deterministic Design Principles
+
+| Principle | Enforcement |
+|-----------|-------------|
+| Three pillars unified | Logs + Metrics + Traces via single SDK; always correlated via trace_id |
+| Vendor-agnostic | OpenTelemetry SDK only; never proprietary SDK |
+| Fixed sampling rates | Dev: 100%, Staging: 50%, Prod: 1-10% |
+| Auto-instrumentation first | Use auto-instrumentation; add manual spans only for business logic |
+| Fixed provider mapping | Datadog, Grafana (OTLP), Sentry — fixed exporter packages |
+| SERVICE_NAME required | Always set via environment variable |
+| No sensitive data in spans | Sanitize all tag values before export |
+
+---
+
+## 9. State & Idempotency Model
+
+Stateless. Fully idempotent. No persistent state.
+
+---
+
+## 10. Failure Handling Strategy
+
+| Failure Class | Behavior | Caller Recovery |
+|---------------|----------|-----------------|
+| Unknown runtime | Return `ERR_UNKNOWN_RUNTIME` | Specify nodejs or python |
+| Unknown environment | Return `ERR_UNKNOWN_ENVIRONMENT` | Specify dev, staging, or production |
+| Unknown provider | Return `ERR_UNKNOWN_PROVIDER` | Specify datadog, grafana, or sentry |
+| Invalid request type | Return `ERR_INVALID_REQUEST_TYPE` | Use supported type |
+
+**Invariant:** Every failure returns a structured error. No partial recommendations.
+
+---
+
+## 11. Error Taxonomy
+
+| Code | Category | Recoverable | Description |
+|------|----------|-------------|-------------|
+| `ERR_INVALID_REQUEST_TYPE` | Validation | No | Request type not supported |
+| `ERR_UNKNOWN_RUNTIME` | Validation | Yes | Runtime not nodejs or python |
+| `ERR_UNKNOWN_ENVIRONMENT` | Validation | Yes | Environment not recognized |
+| `ERR_UNKNOWN_PROVIDER` | Validation | Yes | Provider not one of 3 |
+
+---
+
+## 12. Timeout & Retry Policy
+
+| Parameter | Default | Maximum | Rationale |
+|-----------|---------|---------|-----------|
+| Decision generation | N/A | N/A | Synchronous; < 50ms |
+| Internal retries | Zero | Zero | Deterministic output |
+
+---
+
+## 13. Observability & Logging Schema
+
+### Log Entry Format
+
+```json
+{
+  "trace_id": "uuid",
+  "skill_name": "observability",
+  "contract_version": "2.0.0",
+  "execution_id": "uuid",
+  "timestamp": "ISO-8601",
+  "request_type": "string",
+  "runtime": "string",
+  "environment": "string",
+  "provider": "string|null",
+  "sampling_rate": "number|null",
+  "status": "success|error",
+  "error_code": "string|null",
+  "duration_ms": "number"
+}
+```
+
+### Required Log Points
+
+| Event | Log Level | Fields |
+|-------|-----------|--------|
+| SDK config generated | INFO | runtime, service_name |
+| Sampling rate selected | INFO | environment, rate_percent |
+| Provider recommended | INFO | provider, exporter_package |
+| Instrumentation listed | INFO | libraries_count |
+| Decision failed | ERROR | error_code, message |
+
+### Metrics
+
+| Metric | Type | Unit |
+|--------|------|------|
+| `observability.decision.duration` | Histogram | ms |
+| `observability.runtime.distribution` | Counter | nodejs vs python |
+| `observability.environment.distribution` | Counter | dev vs staging vs prod |
+| `observability.provider.distribution` | Counter | per provider |
+
+---
+
+## 14. Security & Trust Model
+
+### Data Handling
+
+- Observability processes no credentials, API keys, or PII.
+- SDK configuration snippets contain no secrets (SERVICE_NAME via env var).
+- No network calls, no file access.
+
+### Telemetry Security Guidance
+
+| Rule | Enforcement |
+|------|-------------|
+| No sensitive data in span tags | Sanitize before export |
+| SERVICE_NAME via environment variable | Never hardcoded |
+| OTEL_EXPORTER_OTLP_ENDPOINT via env var | Never in code |
+| Provider API keys via env vars | Never in config snippets |
+
+---
+
+## 15. Scalability Model
+
+| Dimension | Constraint | Mitigation |
+|-----------|-----------|------------|
+| Throughput | CPU-bound decision tree | < 50ms; scales linearly |
+| Concurrency | Stateless invocations | Unlimited parallel |
+| Memory per invocation | < 1 MB | No accumulation |
+| Network | Zero network calls | No external dependency |
+
+---
+
+## 16. Concurrency Model
+
+Fully parallel. No shared state. No coordination required.
+
+---
+
+## 17. Resource Lifecycle Management
+
+All resources scoped to invocation. No persistent handles.
+
+---
+
+## 18. Performance Constraints
+
+| Operation | P50 Target | P99 Target | Hard Limit |
+|-----------|-----------|-----------|------------|
+| SDK config generation | < 2 ms | < 5 ms | 20 ms |
+| Sampling decision | < 1 ms | < 3 ms | 10 ms |
+| Full guide | < 10 ms | < 30 ms | 50 ms |
+| Output size | ≤ 2,000 chars | ≤ 5,000 chars | 8,000 chars |
+
+---
+
+## 19. Operational Risks
+
+| Risk | Likelihood | Impact | Mitigation |
+|------|-----------|--------|------------|
+| OpenTelemetry SDK breaking changes | Low | Config snippets outdated | Track OTel releases |
+| Provider exporter deprecation | Medium | Wrong package name | Review annually |
+| Sampling rate too high in prod | High | Excessive cost | Fixed 1-10% enforced |
+| Auto-instrumentation coverage gaps | Low | Missing spans | Manual spans for business logic |
+| High cardinality metrics | Medium | Storage explosion | Reduce label values; use histograms |
+
+---
+
+## 20. Compliance with skill-design-guide.md
+
+| Requirement | Status | Evidence |
+|-------------|--------|----------|
+| YAML frontmatter complete | ✅ | name, description, metadata with category, version, triggers, coordinates_with, success_metrics |
+| SKILL.md < 200 lines | ✅ | Entry point under 200 lines |
+| Prerequisites documented | ✅ | OpenTelemetry SDK packages listed |
+| When to Use section | ✅ | Situation-based routing table |
+| Core content matches skill type | ✅ | Expert type: decision trees, configuration guidance |
+| Troubleshooting section | ✅ | Troubleshooting table with 4 solutions |
+| Related section | ✅ | Cross-links to server-ops, cicd-pipeline, /monitor |
+| Content Map for multi-file | ✅ | Link to engineering-spec.md |
+| Contract versioning | ✅ | contract_version, backward_compatibility, breaking_changes |
+| Compliance matrix structured | ✅ | This table with ✅/❌ + evidence |
+
+---
+
+## 21. Production Readiness Checklist
+
+| Category | Check | Status |
+|----------|-------|--------|
+| **Functionality** | Three pillars (logs, metrics, traces) unified via OTel | ✅ |
+| **Functionality** | SDK setup for Node.js and Python | ✅ |
+| **Functionality** | Fixed sampling (100% / 50% / 1-10%) | ✅ |
+| **Functionality** | Auto-instrumentation (5+ libraries) | ✅ |
+| **Functionality** | Provider routing (3 providers) | ✅ |
+| **Contracts** | Input/output/error schemas in pseudo-schema format | ✅ |
+| **Contracts** | Contract versioning with semver | ✅ |
+| **Failure** | Error taxonomy with 4 categorized codes | ✅ |
+| **Failure** | Zero internal retries | ✅ |
+| **Determinism** | Fixed sampling rates, fixed provider map, fixed SDK config | ✅ |
+| **Security** | No credentials in snippets; all via env vars | ✅ |
+| **Observability** | Structured log schema with 5 mandatory fields | ✅ |
+| **Observability** | 4 metrics defined | ✅ |
+| **Performance** | P50/P99 targets for all operations | ✅ |
+| **Scalability** | Stateless; unlimited parallel | ✅ |
+| **Compliance** | All skill-design-guide.md sections mapped with evidence | ✅ |
+
+---
+
+⚡ PikaKit v3.9.134