Canon

Helix Core Ethos – Runbook v1.0

Helix Core Ethos – Runbook v1.0

Purpose‑Built for the “Two‑Party Approval Flow (TPAF)”
All steps are deterministic, auditable, and require explicit human confirmation before any irreversible action is taken.


Model Governance Overview

See: Helix Models Overview for a comparison of model capabilities, parameter sizes, and approved trust levels.

1️⃣ Overview

{| class=\"wikitable\" | | Item | Description | |---|---| | Runbook ID | HCE‑TPAF‑RB‑001 | | Version | 1.0 (2025‑09‑15) | | Effective Date | 2025‑09‑15 | | Owner | Helix Core Ethos – Governance Team | | Stakeholders | Product Ops, Security, Compliance, Engineering, Legal, End‑User Support | | Scope | End‑to‑end execution of a high‑impact intent (e.g., data‑model training, fund transfer, policy change) that requires two‑party (requester + approver) consent before any irreversible effect occurs. | | Related Docs | • Helix Core Ethos – Principles (TTD‑first, Consent & Least‑Privilege)
• API Spec – intent.json‑ld
• Smart‑Contract ABI – TPAF.sol
• Data‑Retention Profiles (see Appendix A) | | Assumptions | • All participants have an HSM‑backed signing key registered on the Besu ledger.
• Network connectivity to the Besu node is verified (health‑check passed).
• The “Intent” payload is human‑readable and stored in an immutable object store (e.g., IPFS, S3 with Object‑Lock). | | Exclusions | Routine low‑impact intents (Tier 0) that do not require two‑party approval. |


2️⃣ Roles & Responsibilities

{| class=\"wikitable\" | | Role | Primary Responsibilities | Required Access | |---|---|---| | Requester | Drafts the intent, signs the initial payload, submits to the approval queue. | REQUESTER key, read‑only ledger view. | | Approver | Reviews intent, validates compliance, signs to grant consent. | APPROVER key, write‑access to ledger for approval transaction. | | Ops Engineer | Executes the runbook, monitors ledger inclusion, triggers downstream actions after confirmation. | Full node admin, CI/CD pipeline trigger. | | Compliance Officer | Verifies that the intent satisfies all regulatory & policy checks before approval. | Read‑only audit logs, policy‑engine access. | | Security Analyst | Confirms that HSM keys are healthy, validates cryptographic material, monitors for anomalies. | HSM admin, security monitoring dashboard. | | Incident Commander (on‑call) | Handles any failure, escalates per the Incident Response Playbook. | Full system access. |

All actions are logged with a tamper‑evident receipt (Merkle proof) that is automatically emailed to the requester, approver, and compliance officer.


3️⃣ Prerequisites

; Infrastructure

; Credentials

; Software

; Documentation

; Human Confirmation


4️⃣ Procedure – Step-by-Step

All steps are to be executed in order. Any deviation must be logged in Section 7 (Deviation Log).

4.1 Intent Creation (Requester)

Step-by-step actions and expected outputs (see full runbook for exact CLI commands and validations).
Verify requester signature → Run checklist → Human confirmation → Sign and archive.

4.3 Ledger Transaction (Ops Engineer)

Submit transaction → Verify receipt → Email parties → Invoke external irreversible API only after all checks.

4.4 Post-Execution: Clean-up & Retention

Archive artifacts → Log retention profile → Schedule purge if applicable → Close run in dashboard.

5️⃣ Monitoring & Alerts

{| class=\"wikitable\" | | Metric | Threshold | Alert Destination | Description | |---|---|---|---| | Ledger inclusion latency | > 30s/block | Ops Slack #tpa-alerts | Indicates congestion | | Signature failures | ≥ 1 per run | PagerDuty | Potential key compromise | | Email delivery failure | > 0 | Compliance email queue | Must resolve before continuing | | Vault storage error | Any | Ops Email | Prevents data loss | | Unauthorized access | Any | Security SOC | Detects audit breach |


6️⃣ Failure Handling & Rollback

{| class=\"wikitable\" | | Failure Point | Immediate Action | Escalation Path | Evidence to Capture | |---|---|---|---| | Schema validation | Abort, return to Requester | None | Validation log | | Signature failure | Halt, flag for rotation | Security → IC | HSM audit log | | Checklist ❌ | Reject, notify requester | Compliance → Legal | Checklist snapshot | | Ledger not mined | Retry x3, then mark Failed | Ops → IC | Tx hash, node logs | | Email bounce | Re-send alternate channel | Ops → Compliance | SMTP logs | | External API error | Stop, notify parties | IC → Product | API response |


7️⃣ Deviation Log

{| class=\"wikitable\" | | Run ID | Deviation | Reason | Approved By | Date | Mitigation | |---|---|---|---|---|---| | uuid‑example‑001 | Increased gas to 250k | Congestion | Security Lead | 2025‑09‑20 | Script updated |

All deviations must be recorded before a run is closed.


8️⃣ Auditing & Evidence

{| class=\"wikitable\" | | Artefact | Location | Retention | Access Control | |---|---|---|---| | Intent JSON‑LD | S3 (Object-Lock) | 5 yrs | Compliance, Ops | | Signed Package | S3 (Object-Lock) | 5 yrs | Same | | Receipt | Besu archive + Vault | 7 yrs | Auditors | | Emails | Mail archive | 7 yrs | Legal | | Audit logs | PostgreSQL DB | 10 yrs | RBAC (Auditor) |

All artefacts are hash-linked (Merkle root on-chain).


9️⃣ Post-mortem & Continuous Improvement


🔟 Appendices

Appendix A – Data Retention Profiles

{| class=\"wikitable\" | | Profile | Description | Min Retention | Max Retention | Delete Trigger | |---|---|---|---|---| | low-impact | Non-critical metadata | 30 days | 1 yr | Manual | | medium-impact | Analytics results | 180 days | 3 yrs | Auto | | high-impact | Weights, transfers | 5 yrs | 7 yrs | Auto unless legal hold | | regulatory-hold | Legal/subpoena | Indefinite | Indefinite | Legal override |

Appendix B – Compliance Checklist

All items must be ✅:

Purpose Limitation

Data Minimization

Consent Verification

Regulatory Mapping

Risk Severity Rating

Retention Match

Security Review

Audit Trail Integrity

Appendix C – Sample Receipt (JSON-LD)


{
  \"@context\": \"https://schema.org/\",
  \"@type\": \"Receipt\",
  \"identifier\": \"0xabc123…\",
  \"dateIssued\": \"2025-09-15T14:23:07Z\",
  \"transactionHash\": \"0xabc123…\",
  \"requester\": \"req-key-7f9c‑…\",
  \"approver\": \"app-key-3a1e‑…\",
  \"signatures\": {
    \"requester\": \"MEUCIQ…\",
    \"approver\": \"MEUCIQ…\"
  },
  \"merkleProof\": {
    \"root\": \"0xdef456…\",
    \"path\": [\"0x01\", \"0x23\", \"0x45\"]
  },
  \"actionId\": \"ACT‑2025‑00057\",
  \"status\": \"completed\"
}


✅ Final Checklist Before Closing


Prepared by: Helix Core Ethos – Governance Team
Reviewed & Approved: Stephen Hope, Director – 2025‑10‑08

All actions in this runbook are governed by Helix Core Ethos: no hidden training, no dark patterns, no unverifiable claims, and no irreversible actions without confirmed human consent.

Category:Runbooks
Category:TPAF
Category:AI Governance
Category:Operational AI