# label-lab: public projection of local reproduction

This is a public projection of existing execution evidence, not a new run or byte-identical raw output. Named placeholders replace machine-specific paths. Runtime version/OS, command meanings, exits and results are unchanged. The original evidence records executed teaching code, not a cloud task, provider reliability test or model trial.

Runtime requested: `/usr/bin/python3`.
Runtime resolved: `<LOCAL_PYTHON_EXECUTABLE>`.
Version: Python 3.9.6; macOS 26.5.2 arm64. No packages were installed.
Execution window: 2026-10-21T22:28:34.625215+00:00 to 2026-10-21T22:28:38.003998+00:00.

The same `label-lab.zip` was extracted into two new temporary directories. Both were empty before extraction and had no catalog. Source hashes and path-rebased stdout/stderr are in the [public command results](./local-results.public.json). `<LOCAL_PYTHON_EXECUTABLE>` denotes the recorded interpreter installation; `<COPY_A_DIR>` and `<COPY_B_DIR>` denote the original empty extraction roots. The other declared placeholders identify disposable absent-catalog test directories. Both directories were removed after their commands completed.

| Copy | Exact arguments after the runtime | Observed exit and result |
| --- | --- | --- |
| A | `scripts/wrong_prepare.py` | 0: bounded counterexample observed foreground wait; build not reached, catalog absent, own child stopped with exit 0 |
| A | `scripts/lifecycle.py --timeout 0.5` | 1: health 503, catalog unavailable; own child stopped with exit 0; port closed |
| A | `scripts/prepare.py` | 0: build/check plus valid, duplicate-ID and absent-catalog tests pass; preparation returns |
| A | `scripts/lifecycle.py` | 0: health200/ok, catalog200/exact62bytes; own child stopped with exit0; port closed |
| B | `scripts/prepare.py` | 0: independent fresh extraction produces same artifact and passes same3behavior tests |
| B | `scripts/lifecycle.py` | 0: health200/ok, catalog200/exact62bytes; own child stopped with exit0; port closed |

Each preparation actually invoked the converter on duplicate `alpha`: exit2, `duplicate id: alpha`, no output created. Each invoked the checker on an absent path: exit1 with a missing-catalog error. These are expected rejections, not failed positive builds. Their real subprocess outcomes appear in test stdout.

Actual catalog: `observed/catalog.json`, 62 bytes including one final LF.
SHA-256: `b82daf83bcd381d0dc86acf390f89197677e4988441b487d0ab45f2ae3759b90`.
Archive SHA-256: `ef6a9608897cbe4edd4d16f2efc20b62de5d0c7ce227b605ec580675d5d80a61`.

The first ordinary-sandbox attempt failed when the server tried binding127.0.0.1: `PermissionError: [Errno1] Operation not permitted`. That attempt's server exited before the ordering demonstration; no successful readiness was inferred. The `initialAttempt` object in the public command results preserves it. The later successful local runs had separate authorization permitting loopback access. Only owned child processes were controlled; ports were OS-assigned, not taken from another process. No external HTTP service was contacted by the project.

## Repeat from the download

Choose two directory names that do not already exist. From the folder containing the archive:

```sh
python3 -c "from pathlib import Path; Path('fresh-a').mkdir(); Path('fresh-b').mkdir()"
python3 -m zipfile -e label-lab.zip fresh-a
python3 fresh-a/label-lab/scripts/wrong_prepare.py
python3 fresh-a/label-lab/scripts/prepare.py
python3 fresh-a/label-lab/scripts/lifecycle.py
python3 -m zipfile -e label-lab.zip fresh-b
python3 fresh-b/label-lab/scripts/prepare.py
python3 fresh-b/label-lab/scripts/lifecycle.py
```

Continue only after expected exits: the bounded wrong-order demonstration exits0 for observing its intended error; positive preparation/lifecycle also exit0. An absent-catalog lifecycle is separately expected to exit1. The command helpers derive their project root from their own file paths.

The two-copy result covers this exact source archive and one local runtime. Cloud environment creation, publishing, new-task behavior, cloud runtime availability and cloud readiness were not run. F1 is a documentation mapping only. The individual D3 hanging report remains unverified and is not diagnosed by this toy counterexample.