# Phase 1.0Z: offline passive batch contract

Status: `PASSIVE_LF_BATCH_CONTRACT_COMPLETE_LIVE_TRANSPORT_BLOCKED`

Date: 2026-07-22

Phase 1.0Z defines and tests a target-free host model only. It creates no
socket, address, sender CLI, target code or artifact. No PS5 action occurred.

## Contract

An already validated Phase-1.0W `SessionPlan` can produce exactly one immutable
outbound batch:

| Window | Exact byte shape | Bound |
|---|---|---|
| `T2_GREETING_AND_HELP` | `help` followed by LF | 5 bytes |
| `T3_ONE_EXACT_PATH` | `stat PATH` LF, then `sum PATH` LF | 1,035 bytes |

The path is the existing normalized absolute-path alphabet, ASCII-encoded and
at most 512 bytes. NUL, CR, IAC (`0xff`), whitespace inside the path and shell
metacharacters cannot enter the batch. The contract retains no target address.
It permits one outbound batch, no retry, no reconnect and no resume.

Plain LF is accepted by both audited official source families. The legacy
v0.7-v0.8 server passes it through; v0.9-v0.19 `libtelnet` also passes plain LF
to `sh_readline`. The model does not select a deployed family.

## Passive receive and completion

The accumulator accepts only bytes already supplied by a synthetic caller. It
does not own a transport or clock. Any incoming IAC is rejected because this
passive contract emits no negotiation and both audited families send none
proactively. That fail-closed rule avoids silently adding Telnet replies or
signal semantics.

Prompt text and remote EOF are never completion events. Only the explicit
`seal_at_hard_deadline(true)` model event can request sealing. This is a
synthetic boundary, not a working timer or proof that a future blocking receive
can be preempted.

Sealing is successful only when:

- T2 contains a complete help listing matching the known v0.7 or v0.19 source
  family fingerprint;
- T3 contains `stat` metadata including size and a non-cryptographic BSD
  rotate-16 `sum` for the one exact allowed path.

Partial output, unknown help, missing stat, missing sum, invalid UTF-8,
unexpected firmware metadata, IAC and a missing deadline event all fail closed.
The sanitized result never proves an exact binary or device behavior.

## Effects and limitations

The offline model has no side effects. A future real T2/T3 session would still
need explicit acceptance of the automatic model/serial/firmware/temperature/
frequency greeting, shell process and connection state. T3 reads one file;
that may cause low-level effects such as atime, cache, accounting and scheduler
activity. Those effects are not claimed absent.

Unresolved blockers remain:

- exact deployed shsrv identity and family;
- a real bounded adapter and preemptible hard deadline;
- single-write/short-write behavior in current shsrv;
- live chunking, disconnect, cleanup and restart behavior on firmware 9.60;
- physical erasure of sensitive bytes from host memory;
- directory durability for exclusive evidence inherited from Phase 1.0X.

## Decision

The offline passive batch contract is complete. Live transport remains blocked
and unauthorized. The next permitted step is an offline fake-adapter
integration that injects the exact batch and synthetic deadline without adding
a socket, address, device client or device action.