docs(llms): setup agents documentation

.github/skills/code-review/SKILL.md

@@ -0,0 +1,46 @@
+---
+name: code-review
+description: Review Ontime changes for concrete correctness, architecture, testing, routing, security, and maintenance issues. For GitHub Copilot review.
+---
+
+# Review Ontime changes
+
+Review as a maintainer. Find material defects, not speculative advice, style preferences, or generic checklists. Approve when no material issue remains.
+
+## Load relevant context
+
+Read the full diff, PR description, linked issue, tests, and nearby owning code. Then load only applicable guides:
+
+- Every non-trivial review: [change assessment](../../../docs/agent-guides/change-assessment.md)
+- Module placement, server layers, client state, shared packages: [architecture](../../../docs/agent-guides/architecture.md)
+- Tests or changed behaviour: [testing](../../../docs/agent-guides/testing.md)
+- Comments, abstractions, naming, complexity: [code quality](../../../docs/agent-guides/code-quality.md)
+- Authentication, external input, files, integrations, assets, secrets: [security](../../../docs/agent-guides/security.md)
+- Routes, URLs, websockets, authentication, cookies, presets, assets: [routing and cloud](../../../docs/agent-guides/routing-and-cloud.md)
+- Rundowns, timers, persistence, imports, cache, realtime: [domain invariants](../../../docs/agent-guides/domain-invariants.md)
+- Commands, imports, dependencies, formatting, CI claims: [workflow](../../../docs/agent-guides/workflow.md)
+
+Check a guide and nearby canonical code before citing an Ontime convention. Skip unrelated guides.
+
+## Review order
+
+1. Establish intent and affected runtime surfaces.
+2. Read tests first. Identify claimed behaviour and coverage.
+3. Trace implementation, errors, and state transitions.
+4. Check correctness/data integrity, security, architecture/testability, cloud routing, lifecycle/performance, maintainability.
+5. Verify claimed checks. Never claim unobserved results.
+
+Passing tests do not prove architecture, routing, comments, or error paths. Review changed behaviour only; include existing problems only when the diff worsens or relies on them.
+
+## Findings
+
+Report only concrete, actionable issues. Each finding: tight line range, direct defect, impact and trigger, smallest viable remedy when unclear.
+
+- **P0 — Critical:** data loss, exploitable vulnerability, broadly broken production. Blocks merge.
+- **P1 — High:** likely correctness failure or major supported-deployment regression. Blocks merge.
+- **P2 — Medium:** real edge-case defect, architecture regression, missing business-rule test, stale comment, meaningful maintenance risk. Normally blocks merge.
+- **P3 — Low:** local improvement with limited impact. No subjective style or tool-managed formatting.
+
+Order by priority. Prefer few high-confidence findings. No praise or checklist before findings. If none, say so and note verification gaps or residual risk.
+
+Then give one concise PR-level value/risk/complexity assessment. Never repeat it per finding.