scbe-system-engine
Coordinate SCBE-AETHERMOORE math, automation, and service connectors so multi-agent systems can execute work end-to-end. Use when tasks require SCBE dimensional validation, AI-to-AI workflows, browser
SKILL.mdresearch/sphere-grid/skills/scbe-system-engine/SKILL.md
GitHubSCBE System Engine
Operating Contract
- Preserve canonical SCBE terms and spelling in outputs, even if the user input is noisy.
- Preserve the canonical wall formula
H(d*,R) = R · pi^(phi · d*)unless explicitly overridden. - Treat every proposed formula as untrusted until dimensional analysis and behavior checks pass.
- Prefer concrete artifacts over commentary: patch files, scripts, tests, and chain definitions.
- End substantial tasks with a tri-fold YAML
action_summary. - Respect GeoSeal scope discipline: no unrelated side effects in audit or automation runs.
Scope
Use this skill when tasks involve:
- Math/physics validation (
dimensional-analysis.md,layerinteractions, boundedness checks). - Multi-agent design (StateVector + DecisionRecord outputs, tongue routing).
- Service orchestration (GitHub, Hugging Face, Notion, Linear, Zapier).
- Browser-backed discovery or Playwright capture steps.
- Self-improvement loops that convert observed gaps into patch files.
Workflow
- Identify layer(s), formula(s), and connected services in scope.
- Read references before output:
references/scbe-glossary.mdreferences/scbe-constants.mdreferences/dimensional-analysis.mdreferences/ai-to-ai-comms.md
- Run or draft only deterministic, inspectable outputs (tests, scripts, chain YAML, dataset card metadata).
- Separate
build,document, androutelegs. - Finish with a tri-fold
action_summary.
Required Dual Output
SCBE compliance checks must emit:
StateVector: deterministic technical stateDecisionRecord: action, signature, timestamp, reason, confidence
Tongue and Model Routing
KO→claude-opus-4-6(Engineering/Tight systems checks)AV→claude-sonnet-4RU→claude-opus-4-1CA→claude-opus-4-6(math/crypto/code validation)UM→claude-sonnet-4DR→claude-opus-4-6
SCBE Layer Sequence (Canonical Reference)
- Complex context state
- Realification
- Weighted transform
- Poincaré embedding
- Hyperbolic distance
- Breathing transform
- Phase transform
- Multi-well realms
- Spectral coherence
- Spin coherence
- Triadic temporal
- Harmonic scaling
- Decision + response
- Audio axis
Resources
references/dimensional-analysis.md: formula checks, transform validity, scaling risks.references/service-automation-contract.md: MCP discovery and connector routing.references/browser-playwright-notes.md: browse-safe automation patterns.references/self-improvement-loop.md: observation → diagnosis → patch.references/ai-to-ai-comms.md: typedtool/llm/gatechain schema.
Script Map
scripts/pqcm_audit.py
Property-based stress checks for formula proposals such askappa_eff.scripts/ko_tongue_code_reviewer.py
KO-tongue agent class that returnsAgentOutput.scripts/route_github_linear_chain.yaml
End-to-end branch chain for GitHub PR event, code review, PR comment, and Linear fallback issue.
Service Routing Rules
- Hugging Face, Notion, GitHub, Linear, and Zapier:
- discover tool availability first,
- report
callable nowvsneeds configuration, - never claim capabilities not currently available.
- Do not modify the canonical wall formula unless explicitly required.
Output Contract
Every route request should return:
files_changed: explicit paths.rationale: short explanation of why the patch or behavior is needed.services_to_update: explicit MCP/service identifiers.pending_integrations: unresolved setup tasks.