cto/docs/LEGACY-INGEST.md

# CTO Legacy Ingest

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

Source: CTO-WORK-097.

## Purpose

Legacy Ingest captures intention; it does not import implementation mass.

Use this file to decide which older CTO planning clusters still matter for the current child-local endgoal:

- Cortex governs.
- Hermes controls.
- CTO routes.
- Case executes.
- Harness proves.
- Target repo stays owned.
- Evidence records.
- Core promotes only through SOT route.

## Source Clusters Assessed By Intention

### Case Admission And Proof Contract

Reference cluster:

- `.sot/03-PROTOCOLS/CTO-CASE-SOURCE-ADMISSION-RECORD.md`
- `.sot/03-PROTOCOLS/CTO-CASE-ADAPTER-CONTRACT.md`
- `.sot/03-PROTOCOLS/CTO-HARNESS-EVIDENCE-INTERFACE-CONTRACT.md`
- `.sot/03-PROTOCOLS/CTO-CASE-FAILURE-FIXTURE-MATRIX.md`
- `.sot/03-PROTOCOLS/CTO-CASE-STAGED-PROOF-GATES.md`

Decision: keep the intention.

The useful idea is the proof shape: source admission, adapter boundaries, failure fixtures, staged gates, and evidence contracts. Do not preserve every intermediate scaffold if it only restates the same gate.

### Governed Execution Flow

Reference cluster:

- `CTO-WORK-066` through `CTO-WORK-071`
- `.sot/03-PROTOCOLS/CTO-GOVERNED-EXECUTION-REQUEST-PRD.md`
- `.sot/03-PROTOCOLS/CTO-GOVERNED-EXECUTION-APPROVAL-RECORD.md`
- `.sot/03-PROTOCOLS/CTO-GOVERNED-EXECUTION-EVIDENCE-CLOSEOUT.md`

Decision: keep the intention.

The useful idea is a governed request, explicit approval, execution evidence, and closeout. The target repository remains owned by its own route. CTO does not become a general execution authority.

### Hermes Visibility

Reference cluster:

- `.sot/03-PROTOCOLS/CTO-HERMES-CONTROL-SURFACE-PRD.md`
- `.sot/03-PROTOCOLS/CTO-HERMES-REAL-REFRESH-CONTROL-REPLAY-PRD.md`
- `.sot/03-PROTOCOLS/CTO-HERMES-WEBUI-CONTROL-PANEL-EVIDENCE.md`

Decision: keep the intention.

The useful idea is visibility and replay. Hermes may display proof posture, approval state, and blocked lanes. Hermes must not govern, promote, or mutate target repositories from this child workspace.

### Provider And Retention Work

Reference cluster:

- `.sot/03-PROTOCOLS/CTO-CASE-PROVIDER-DECISION-RECORD.md`
- `.sot/03-PROTOCOLS/CTO-CODEX-RETENTION-DRY-RUN-PACKET.md`
- `.sot/03-PROTOCOLS/CTO-CODEX-RETENTION-POLICY-PACKET.md`
- `.sot/03-PROTOCOLS/CTO-CODEX-RETENTION-ARCHIVE-EXECUTOR-PACKET.md`

Decision: keep the intention with caution.

The useful idea is explicit provider admission and metadata-only retention hygiene. Do not copy secrets, raw transcripts, or provider runtime behavior into CTO. Archive and deletion remain approval-gated.

### Stage 5 Target Sandbox

Reference cluster:

- `/home/svrnty/workspaces/cortex-os/cto-stage5-target-sandbox`
- `.sot/03-PROTOCOLS/CTO-CASE-STAGE5-TARGET-REPOSITORY-ADMISSION.json`
- `.sot/03-PROTOCOLS/CTO-CASE-STAGE5-OWNED-NONCRITICAL-REPO-PRD.md`
- `.sot/03-PROTOCOLS/CTO-CASE-STAGE5-REPEATABILITY-PROOF-EVIDENCE.md`
- `tools/validate_cto_child.py`

Decision: keep the intention; physical custody complete.

The useful idea is an explicitly owned, noncritical, disposable target repository for Stage 5 proof. The original target path is now historical proof context only, and the preserved payload is outside the umbrella.

Current evidence:

- 2026-06-17 refresh: current path is not a Git repository.
- 2026-06-17 refresh: `python3 tools/validate_cto_stage5_target_sandbox_child.py` passes in the target.
- 2026-06-17 refresh: `python3 tools/validate_cto_child.py` passes in CTO while the root path exists.
- 2026-06-17 refresh: `tools/validate_cto_child.py` hard-codes the target path and admission record fields.
- 2026-06-17 refresh: `.case/tasks/active/*.task.json` contains six active task records; two have `agents.implementer.status` set to `running` and no `completed` timestamp.
- 2026-06-17 refresh: the target task records still use active or implementing states, so they are not archive-stable closure evidence.
- 2026-06-17 stale-state declaration: `docs/STAGE5-TARGET-SANDBOX-STALE-CASE-STATE.md` records that target tests and the target validator pass, no live Case or `ca` process is present, and the six target-local Case records are stale residue rather than current work authority.

Movement gate:

2026-06-17 custody preflight: the old Stage 5 target admission is now
historical proof context only. The admission record sets
`admission_lifecycle=historical-proof-target`,
`future_execution_authorized=false`, and
`future_execution_requires_new_admission=true`. Physical relocation remains a
separate host-aware custody gate because the MacBook mirror can rehydrate root
paths unless policy is checked first.

2026-06-17 mirror blocker check: the MacBook source path
`/Users/jean-philippebrule/Steev/Cortex-OS/cto-stage5-target-sandbox` exists,
and `/Users/jean-philippebrule/Steev/push-cortex-os-to-steev-once.sh` does not
exclude `cto-stage5-target-sandbox/`. The custody preflight now records an
archive-stable evidence candidate for the target, but physical movement remains
blocked until mirror policy is updated or JP chooses a different MacBook source
custody policy.

2026-06-17 physical custody result: the MacBook push script now excludes
`cto-stage5-target-sandbox/` with backup
`/Users/jean-philippebrule/Steev/push-cortex-os-to-steev-once.sh.bak.20260617-cto-target-exclude`.
The Steev target moved to
`/home/svrnty/workspaces/cortex-os-retired/2026-06-17/cto-stage5-target-sandbox-preserved-091700`
with `MANIFEST.md`. The preserved target validator passes from the preserved
path, pytest passes with cache provider disabled, and the original root path
stayed absent through the observed MacBook push window ending
`2026-06-17T09:18:23-04:00`.

Current closure: the absolute live path dependency is closed for current
execution authority. Future Stage 5 execution still requires a new admission
record, new approval, current target validation, and fresh Harness evidence.

Rejected now: hard delete, cleanup-only target mutation, Case rerun, Harness rerun, provider mutation, target source mutation, Core mutation, Seed mutation, readiness claim, release claim.

## Rejection Rules

- Do not activate Case as default backend.
- Do not mutate target repositories.
- Do not treat Hermes, Case, Pi, Codex, or any execution backend as Cortex OS authority.
- Do not promote child-local CTO text into Core without a governed Core route.
- Do not ingest duplicate planning drafts when a later validated closeout already captures the intention.

## Completion Stage

Stage: legacy intention captured for the main CTO planning clusters.

Remaining useful work: compress duplicate CTO planning surfaces after each cluster has an owner, validator coverage, and a current route decision.