hermes/cto
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
46b7296cb2
cto/.sot/03-PROTOCOLS/CTO-CASE-STAGE6-REPEATABILITY-REFRESH-PRD.md
Svrnty c0ea3207d1 Record Stage 6 repeatability refresh
2026-06-02 06:48:26 -04:00

5.5 KiB
Raw Blame History

name status lifecycle_classification owner created last_reviewed core_promotion_status source
CTO Case Stage6 Repeatability Refresh PRD validated sot jp 2026-06-02 2026-06-02 not-promoted .sot/03-PROTOCOLS/CTO-CASE-STAGE5-REPEATABILITY-PROOF-EVIDENCE.md

CTO Case Stage6 Repeatability Refresh PRD

Local planning SOT only. Not a Core Protocol. Not active Core authority.

Stage 6 repeatability refresh.

Problem

CTO has two governed Stage 5 Case passes in the admitted Target Repository. Existing Stage 6 real-governed refresh evidence consumed only the first pass. CTO needs a read-only refresh that proves repeatability across both passes before any later candidate-default decision discussion.

Solution

Extend the existing Stage 6 real-governed refresh seam. The seam reads two pinned Harness Evidence Interface report and proof pairs, verifies both governed passes against Stage 6 dimensions, records computed repeatability status, and exposes repeatability through the Hermes Control Surface summary.

Pinned Inputs

  • first case_id: r1-src-string-slugify
  • first run_id: 20260601T105222Z-r1-src-string-slugify-180161
  • first report: /home/svrnty/.hermes/profiles/cto-planb/harness-runs/20260601T105222Z-r1-src-string-slugify-180161/report.json
  • first Stage 5 proof: /home/svrnty/.hermes/profiles/cto-planb/harness-runs/20260601T105222Z-r1-src-string-slugify-180161/stage5-owned-repo-proof.json
  • first target commit: 7706f99
  • first expected changed files: src/strings.py, test_strings.py
  • repeat case_id: r1-slugify-punctuation
  • repeat run_id: 20260602T101450Z-r1-slugify-punctuation-2663598
  • repeat report: /home/svrnty/.hermes/profiles/cto-planb/harness-runs/20260602T101450Z-r1-slugify-punctuation-2663598/report.json
  • repeat Stage 5 proof: /home/svrnty/.hermes/profiles/cto-planb/harness-runs/20260602T101450Z-r1-slugify-punctuation-2663598/stage5-owned-repo-proof.json
  • repeat target commit: 2fa1601
  • repeat expected changed files: src/strings.py, strings.py, test_strings.py

Scope

  • Validate both report/proof pairs.
  • Verify report shape, event validity, allowed-path compliance, failure closure, failure matrix coverage carry-forward, artifact completeness, forbidden-action closure, operator acceptance, approval provenance, source admission freshness, target validation, and Target Repository read-only status.
  • Require fake lane evidence from prior Stage 6.
  • Keep Codex lane blocked with rationale.
  • Keep Pi lane blocked with rationale.
  • Record computed repeatability_status as pass, blocked, or fail.
  • Record repeatability_count: 2 when both pinned passes validate.
  • Record runtime default activation: false.
  • Record core_promotion_status: not-promoted.
  • Expose repeatability count, status, comparison path, and default activation false through Hermes Control Surface summary.

Non-Goals

  • Do not run a new Case mutation.
  • Do not activate Case as default backend.
  • Do not promote CTO artifacts into Core.
  • Do not mutate a Target Repository.
  • Do not push, merge, deploy, close, PR open, issue close, publish, mutate vendor source, mutate external developer repositories, mutate unowned repositories, expose endpoint values, expose secret values, expose credential values, or expose raw Target Repository content.
  • Do not replace Core authority with CTO, Hermes, Harness, Case, Pi, Codex, or any execution backend.

Acceptance Criteria

  • CTO-WORK-081, CTO-WORK-082, CTO-WORK-083, and CTO-WORK-084 are validated.
  • validate-case-stage6-real-refresh.py consumes both pinned report/proof pairs.
  • validate-case-stage6-real-refresh.py records repeatability_count: 2.
  • validate-case-stage6-real-refresh.py records repeatability_status: pass.
  • validate-case-stage6-real-refresh.py records runtime_default_activation: false.
  • validate-case-stage6-real-refresh.py records core_promotion_status: not-promoted.
  • Hermes Control Surface summary records repeatability_count: 2.
  • Hermes Control Surface summary records repeatability_status: pass.
  • Hermes Control Surface summary records candidate_default_repeatability_eligible: true.
  • Hermes Control Surface summary records runtime_default_activation: false.
  • Validators reject missing artifacts, failed pass facts, target mutation, default activation, Core promotion, forbidden actions, missing approval provenance, stale source admission, and secret-shaped keys.

Validation

  • python3 harness/runner/validate-case-stage6-real-refresh.py --json
  • python3 harness/runner/validate-webui-summary.py --json
  • harness/evals/health.sh --json
  • python3 tools/validate_cto_child.py
  • python3 /home/svrnty/workspaces/cortex-os/core/tools/check_s69_caveman_prose_discipline.py

Risks

  • Repeatability evidence can be misread as default authority. Mitigation: record runtime default activation false in every artifact.
  • A second pass can overstate readiness if failure closure is ignored. Mitigation: require prior Stage 6 failure closure.
  • Artifact paths can drift. Mitigation: validators load files from pinned artifact paths.
  • Approval provenance can drift between prose and proof. Mitigation: validator records proof approval_source exactly and evidence records the mismatch.
  • Control Surface wording can imply governance. Mitigation: Core promotion remains not-promoted.

Success Definition

CTO has validated child-local evidence that Case passed two distinct governed Stage 5 tasks through the CTO Harness seam, Hermes exposes that repeatability evidence read-only, runtime default activation remains false, and Core promotion remains not-promoted.