ManPage.md: mlpstorage status and submit — OPTIONS sections, Submission Workflow rewrite, FILES entries for the package and ledger (status-and-submit PR 4)

[FileSystemGuy]

Fourth and last PR of the status / submit series (#879 vocabulary + readiness evaluator, #880 mlpstorage status, #881 mlpstorage submit). Docs only: ManPage.md catches up with what the three code PRs shipped. --help_all already carried the STATUS and SUBMIT blocks (the parity test forced them in with the code), so this PR is the man page.

What changes in ManPage.md

• SYNOPSIS / COMMAND STRUCTURE — mlpstorage status [OPTIONS] and mlpstorage submit [--dry-run] [--out PATH] added beside runs.
• Submission Workflow — rewritten around Rules.md 1.3 (run / result / submission, runs_per_result from editions.yaml). Steps 3–6 are now: run until the result is complete (the post-run status row shows RUNS n/m after every run), fill in the paperwork, mlpstorage status until every result is ready, mlpstorage submit --dry-run then submit. The code/ bullet (stale since the code-image pool) becomes code-images/. validate is named as the reviewer's tool over the same checker.
• OPTIONS → Status (new) — the table (rendered with the real formatter), RUNS / SUBMIT / NOTE / footer semantics, the four result tokens in precedence (short > invalid > paperwork > ready, - whatif), the five per-run tokens, paperwork-vs-rerun as the severity axis, which rule IDs are folded into RUNS and into submit's regeneration step, every flag, the --json shape.
• OPTIONS → Submit (new) — the six steps as built (reportgen first in both modes; one checker pass; refusal text + exit 1; package plan <org>/{closed,open}/<org>/** + referenced images under <org>/code-images/, never whatif/, .mlps/, the sentinel or orphan images; dry-run output; tarball + .sha256 + .manifest.json schema mlps-submission-package/1 with the key list; .mlps/submissions.jsonl event shape; manual upload per Submission_guidelines §11), --dry-run / --out semantics, exit status. States that there is no --force.
• Runs → list — the SUBMIT column and the JSON submit field (- / ? / null cases).
• RESULTS DIRECTORY — .mlps/submissions.jsonl and .mlps/packages/ in the tree; "nothing under .mlps/ is ever packaged".
• VALIDATOR — one sentence each in the intro and Invocation: status / submit run the same checker in-process; one implementation of every rule.
• END-OF-RUN RECAP — the post-run status table (stdout, after metadata + log detach, skipped under --quiet, never changes the exit code) documented next to the stderr recap.
• EXIT STATUS / EXAMPLES / FILES / SEE ALSO — submit refusal = 1, status always 0; a status→submit example; FILES entries for the ledger and the package + manifest; Submission_guidelines §11 pointer.

Tests

tests/unit/test_submit_command.py::TestManPage (6 tests, RED first in the first commit): SYNOPSIS + tree entries, the Status section names every flag and all nine tokens plus runs_per_result and Next:, the Submit section names --dry-run/--out/reportgen/sidecars/ledger/code-images/whatif/--force/exit, FILES has the ledger + package + schema entries, the workflow ends in status and submit, the Runs section mentions SUBMIT.

All four CI suites green locally (tests, mlpstorage_py/tests, vdb_benchmark/tests, kv_cache_benchmark/tests). No code change, no validator change, so no validator-diff run.

Reviewable calls

1. The Submission Workflow bullet list now says code-images/ (Rules.md 2.1.6) instead of code/ with the MD5 sentence. Submission_guidelines.md §11.1 still says code — left alone here (that file is the WG's, and the ManPage points at §11 only for the upload UI).
2. submit is described as recorded in .mlps/history (it is: main.py records every writer; only history, runs, status are skipped).
3. The Status example table is the real formatter's output for a plausible tree; the checkpointing row shows 2/2 invalid (both phases counted, plus a failed run to remove), which is how readiness scores that case.

[github-actions]

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅