Gladex Agent Logs

      "0.2",
      "0.3",
      "0.4"
    ],
    "ok": true,
    "reason": null,
    "summary": "116 changelog version heading(s), 116 unique, 3510 citation(s) checked, 0 missing"
  },
  "ok": true,
  "exit_code": 0
}
rc=0
{"changelog": "/data/repo/CHANGELOG.md", "detail": "one queue: [0.4.111] pointer-only, 111 item line(s) frozen across 68 section(s), 72 PROGRESS bullet(s)", "error": null, "exit_code": 0, "ok": true, "progress": "/data/repo/agent-logs/PROGRESS.md", "repo_root": "/data/repo", "rules": {"duplicate_queue_sections": 0, "frozen_items_actual": 111, "frozen_items_expected": 111, "newest_has_queue_section": true, "newest_points_at_progress": true, "newest_queue_items": 0, "newest_version": "0.4.111", "progress_bullets": 72, "queue_headings_fenced": 1, "queue_headings_raw": 69, "queue_sections": 68}, "timestamp": "2026-09-28T08:44:40Z", "tool": "queue-source-check", "violations": []}
rc=0
> build · mimo-v2.6-flash-free
$ cd /data/repo && cat >> CHANGELOG.md <<'EOF'

## [0.4.112] - 2026-09-28 — queue item (59): the promote-gates contract had three prose copies and the tool's own "pinned by this suite" sentence had no reader — new section 20 compares all three, suite 295 → 323

### Why

Same failure mode as `[0.4.108]` (go-tests) and `[0.4.111]` (go-compile), one child further along: `tools/system-status` says `# Contract (pinned by tests/test_system_status_promote_gates.sh):` above the `promote-gates` row, and **nothing in that suite read the sentence**. Measured per child by (57): the suite reads the tool's source in sections 15/15b/15c/18/18b, but only for invocation strings, env tokens and mailbox stamps — so the `promote-gates verdicts` block in `--help`, the contract comment, and this suite's own `# The contract under test:` block were three hand-maintained renderings of one contract with **no assertion comparing any two of them**, and a one-sided edit of any copy changed nothing the suite could see.

### Changed

- **`tests/test_system_status_promote_gates.sh` gained section 20 (28 assertions, suite 295 → 323).** It extracts all three copies — the `--help` verdict block (aligned two-column register), the tool's contract comment and this suite's header (both `left -> verdict` prose with wrapped conditions) — and compares them as a **(declared condition key → verdict)** map of **3 keys declared from promote's own vocabulary**: `all-pass`, `refused`, `unverifiable`. This contract carries **no exit codes anywhere**, which is exactly why go-compile's `rc_class()` key table could be reused in *shape* but not in content; each key accepts both registers' phrasing of the same condition (`all four gates pass` vs `'promote-ready' verdict, dev-sync, commit-lint AND ship-tree all pass`). An arrow whose right side does not begin with a verdict word is prose, not a case — the comment's three divergence-guard branches (`child_phrase()`, `HEAD moved`, `the verdicts cannot be compared`) are skipped rather than read as a fourth case.
- **Non-vacuity asserted, not assumed**: each copy must land on exactly **3** entries carrying **3 distinct keys**; `spg_equal` refuses two empty extractions **and that refusal is itself asserted** alongside its acceptance of the three real equal maps; and **no copy may map any promote case to `error`** — the row's whole promise is that a refused promotion is not a system failure (the tool stays exit 0), so `refused|error` in any register is a red regardless of whether the copies agree with each other.
- **Six negative controls — a verdict flip and a key deletion per copy.** Each plant first asserts it changed the text it came from, so "caught" cannot be true because the substitution matched nothing. The three deletions are caught by the map **shrinking 3 → 2** rather than by an inequality alone: renaming `--help`'s `'cannot verify'` label (its row's only key carrier), rewording the comment's `all four gates pass`, and rewording the header's `everything we could NOT check`.
- **Claim → reader**: the tool's `pinned by tests/test_system_status_promote_gates.sh` sentence is asserted, and the block it names is asserted to be non-empty and to carry its second case — the two lines that make the claim true rather than merely present.
- **A seam declared, not papered over**: a rewording that *keeps* a copy's key (`a gate refuses` → `a gate said no`) stays green across registers — the compare-or-declare choice already parked as (54)/(56), recorded in the section's own comment instead of hidden behind a comparison that would be red on wording nobody thinks is a copy.

### Verified

- **Three live replays against the real files**, not just strings extracted into shell variables: planting the drift in the tool's **comment** → `318 passed / 5 failed`, in the tool's **`--help` block** → `318 / 5`, and in **this suite's own header** → `319 / 4` — with **every red inside section 20** each time (the other suites' sections stayed green because they assert behaviour, which the plant does not change), and each file restored `md5`-verified (`tools/system-status` and the suite both back to their pre-plant digest). In each replay the affected copy's own N-control also went red — its "plant must differ from the text" guard correctly refusing to call a no-op caught.
- **`tools/regression-run` → 57 suites, 5024 assertions, 2 failed** (pre-commit; `4996 + 28 = 5024`, the arithmetic that says the step added only itself). The 2 reds are `test_gladex_monitor.sh`'s dirty-tree pair `A3`/`A15` (`git status` sees this run's own uncommitted files), the pair documented as red-before-commit/green-after in `[0.4.107]`.
- `tools/repo-lint --format json` → **exit 0, `ok true`, `failures []`, `duplicates []`, `citations_missing []`, `go_compile.ok true`**, 116 entries before this one; `tools/queue-source-check --format json` → **exit 0, `violations []`**, `[0.4.111]` pointer-only, **111** frozen item lines, `newest_queue_items 0`; `bash tests/test_system_status_promote_gates.sh` → **323 / 0 / 0**.

### Notes

Only the suite changed in `tests/`; `tools/system-status` is **byte-unchanged** — both of its copies were already correct and already agreed, which is the whole point of (57): the readers were missing, not the text. `tools/REGISTRY.md`'s promote-gates bullet carries the new count and the section-20 description. Item **(59)** struck; **(60)** and **(54)** (the `REGISTRY.md` evidence table / exit-code bullets — compare or declare) remain open.

### Queue -> agent-logs/PROGRESS.md
Authoritative next-candidate queue lives in `agent-logs/PROGRESS.md`, in the `Next-candidate queued, not actioned` bullet of the newest run entry — it is not duplicated here, and has not been since `[0.4.67]`. The 111 item lines already written into the historical `### Queue` sections are frozen history. To read the live queue, open `PROGRESS.md`; to know whether that is still true, run `tools/queue-source-check`.
EOF
grep -c '^## \[' CHANGELOG.md; grep -n '^## \[0.4.112\]' CHANGELOG.md; tail -3 CHANGELOG.md | head -1
117
4093:## [0.4.112] - 2026-09-28 — queue item (59): the promote-gates contract had three prose copies and the tool's own "pinned by this suite" sentence had no reader — new section 20 compares all three, suite 295 → 323
> build · mimo-v2.6-flash-free
$ cd /data/repo && ./tools/repo-lint --format json | python3 -c 'import json,sys; d=json.load(sys.stdin); print("repo-lint", d["exit_code"], d["ok"], d["failures"], d["changelog_version"]["summary"])'; ./tools/queue-source-check --format json | python3 -c 'import json,sys; d=json.load(sys.stdin); print("queue", d["exit_code"], d["ok"], d["violations"], d["rules"])'
repo-lint 0 True [] 116 changelog version heading(s), 116 unique, 3510 citation(s) checked, 0 missing
queue 0 True [] {'duplicate_queue_sections': 0, 'frozen_items_actual': 111, 'frozen_items_expected': 111, 'newest_has_queue_section': True, 'newest_points_at_progress': True, 'newest_queue_items': 0, 'newest_version': '0.4.112', 'progress_bullets': 72, 'queue_headings_fenced': 1, 'queue_headings_raw': 70, 'queue_sections': 69}
$ cd /data/repo && for t in test_repo_lint.sh test_queue_source.sh; do printf "%s: " "$t"; bash tests/$t 2>&1 | tail -1; done; for t in test_changelog_api.php test_changelog_mobile.php test_app_version.php test_cli_version.php; do printf "%s: " "$t"; php tests/$t 2>&1 | tail -1; done
test_repo_lint.sh: ===============================
test_queue_source.sh: ===============================
test_changelog_api.php: === Results: 86 passed, 0 failed ===
test_changelog_mobile.php: === Results: 125 passed, 0 failed, 0 skipped ===
test_app_version.php: === Results: 39 passed, 0 failed ===
test_cli_version.php: === Results: 35 passed, 0 failed ===
> build · mimo-v2.6-flash-free
$ cd /data/repo && grep -n "CHANGELOG" tools/repo-lint | head -20
45:CHANGELOG version-identity gate (why a test is not a gate):
47:    several identities append to CHANGELOG.md, so a remembered rule cannot
54:    nothing at run time checked the commit. So the COMMITTED CHANGELOG.md
59:    the changelog API skips such headings too. A tree with no CHANGELOG.md
63:    committed file that cites `[x.y.z]` in a series this CHANGELOG has an
78:    scope too - and with no CHANGELOG.md to compare against, the tokens are
110:        CHANGELOG.md)
112:        not compile, or the committed CHANGELOG.md repeats a version token,
113:        or a committed file cites a version CHANGELOG.md never wrote
132:compile and CHANGELOG version gates existed - while five more sections
222:# CHANGELOG entry heading: `## [1.2.3]` - the version token /api/changelog
227:CHANGELOG_PATH = "CHANGELOG.md"
228:CHANGELOG_HEADING_RE = re.compile(r"^## \[(\d+\.\d+\.\d+)\]")
235:CHANGELOG_CITATION_RE = re.compile(rb"\[(\d+\.\d+\.\d+)\]")
255:        for m in CHANGELOG_CITATION_RE.finditer(line):
272:    compile and CHANGELOG version gates existed) while the epilog already
302:     "CHANGELOG version-identity gate (why a test is not a gate):"),
916:    """Version identity of the COMMITTED CHANGELOG.md - see docstring.
931:    Absence is NEITHER: a tree that ships no CHANGELOG.md has nothing to be
958:            {"path": CHANGELOG_PATH, "lang": "changelog", "error": message}
> build · mimo-v2.6-flash-free
$ cd /data/repo && bash tests/test_repo_lint.sh 2>&1 | grep -E "Results|passed|FAIL" | tail -5; echo "rc=$?"; bash tests/test_queue_source.sh 2>&1 | grep -E "Results|passed|FAIL" | tail -5
PASS: V1: ...so a reader never reconciles 2 FAIL lines against 2 file(s)
PASS: W3: human FAIL line still names the file
PASS: X3: the FAIL line it refers to is still printed above it
PASS: M20 planted: the derivation is defeated, not bypassed
repo-lint suite: 420 passed, 0 failed
rc=0
queue-source-check suite: 181 passed, 0 failed
> build · mimo-v2.6-flash-free
$ cd /data/repo && grep -n "Next-candidate\|progress_bullets\|authoritative" tools/queue-source-check | head -30
8:    `Next-candidate queued` bullet - two lists, hand-maintained, in two files.
17:    `agent-logs/PROGRESS.md` is the SINGLE authoritative next-candidate queue.
44:    R5  agent-logs/PROGRESS.md carries at least one `Next-candidate queued`
45:        bullet - the authoritative list still exists
61:What R5 COUNTS (it is reported as `progress_bullets`, and until [0.4.80] that
62:    number was `pg_text.count("Next-candidate queued")` - raw occurrences
67:    without a list marker (`- **Next-candidate queued…**` and the six entries
68:    written as a bare `**Next-candidate queued…**` paragraph both count, once
129:BULLET_START_RE = re.compile(r"^[ \t]*(?:[-*+][ \t]+)?(?:\*\*)?Next-candidate queued")
133:    """How many lines OPEN a `Next-candidate queued` entry - never occurrences.
185:      authoritative list is empty, one entry owns two `### Queue` sections, or
292:            "progress_bullets": None,
339:    # A3: the authoritative file must be readable before its content is judged.
382:    # R5: the authoritative list exists.
384:    report["rules"]["progress_bullets"] = bullets
387:            "%s carries no 'Next-candidate queued' bullet - the authoritative "
485:            "progress_bullets": None,
> build · mimo-v2.6-flash-free
$ date -u +"%Y-%m-%dT%H:%MZ"; cd /data/repo && git log --oneline -3 && git status --porcelain
2026-09-28T08:48Z
2c4e750 docs: post-commit confirmation — regression 57/4996/0 'all suites green' twice, repo-lint rc 0 (116 entries, no dups), monitor 27/0 with A3+A9+A10 green, STEP 0 re-checked 0/0 at close, tree clean at 81874e4
81874e4 docs: [0.4.111] + PROGRESS — (57) and (58)(a) executed: go-compile contract copies compared (82 → 109), QA entry re-homed to [0.4.110], regression 57/4996/0 "all suites green"
1737e69 fix: re-home the prepended QA entry — [0.4.29] (2026-09-28) → [0.4.110] parked at the bottom, `# Changelog` back on line 1
 M CHANGELOG.md
 M tests/test_system_status_promote_gates.sh
 M tools/REGISTRY.md
> build · mimo-v2.6-flash-free
$ cd /data/repo && cat >> agent-logs/PROGRESS.md <<'EOF'

## 2026-09-28T08:48Z main-loop run — STEP 0 answered FIRST (0 unread both DBs; fresh-probe reply **dev 105 / prod 71**), then ONE step: **(59) EXECUTED** — section **20** compares the promote-gates contract's three prose copies, suite **295 → 323**

- **STEP 0 (first action, before any product work)**: unread `investor_to_agent` = **0 dev / 0 prod**, read with a direct `SELECT count(*) FROM messages WHERE direction='investor_to_agent' AND read=0` on both live SQLite DBs (`/opt/startup/{dev,prod}/data/messages.db`), so nothing was marked read that is not mine. A fresh-probe reply was inserted **parameterised** (python3 `?` binding, body held in a script variable, never interpolated into shell or printed; the script deleted immediately after the insert) → **dev 105 / prod 71**, `read=1`; re-verified after the insert that the newest row in each is `agent_to_investor` / `read=1` and unread is still **0 / 0**. `tools/inbox-status` → **exit 0, `owed.total` 0**, 74 entries / 73 handled / **1 open** with `open_unreplied []` — line 468, the 2026-09-26 *"agents get matching accounts on Nextcloud + Immich"*: its **Immich half is delivered** (8 accounts, every identity logging in, admin-only routes 200 for admin / 403-401 for the rest), its **Nextcloud half is still blocked** on REPORT.md §14 (`https://cloud.gladex.de/status.php` → `{"installed":false}`), and the reply restates the one line that closes it — *does `password = own email` extend to Nextcloud?* — plus what will be done the moment it arrives (eight admin-created accounts, never self-registered, a real login probe per account, names-only matrix in `STRUCTURE.md`, no password in git/logs/thread/file). The reply also names this run's step, so the block is actionable from the investor's side rather than only reported.
- **The step, and the defect inside it**: **(59)**, queued by the 07:17Z entry as *"the same unread-comment gap exists for promote-gates: `tools/system-status` L1329's `# Contract (pinned by tests/test_system_status_promote_gates.sh):` is never grepped by that suite"*. Re-measured this run, per child rather than assumed: this suite **does** read the tool's source (`SRC=$([ -n "${SYSTEM_STATUS_BIN:-}" ] && cat "$TOOL" || cat …)`, L1227) but only with `grep -F` for the invocation line, `GLADEX_GATE_PROBE=1`, the default child path and the mailbox stamps — never for the contract sentence, and never for `--help`'s `promote-gates verdicts` block or for this file's own `# The contract under test:` block (L10–L84). Three hand-maintained renderings of one contract, **zero assertions comparing any two**, so a one-sided edit of any copy was invisible by construction — the same shape as (55) on go-tests and (57) on go-compile, one child further along.
- **Why the key table had to be rebuilt, not reused — this is the caveat (59) itself named**: this contract carries **no exit codes anywhere**, so go-compile's `rc_class()` table (`clean`/`child-nonzero`/`broken`/`unverifiable`) has nothing to key on. The three keys come from promote's own vocabulary — **`all-pass` / `refused` / `unverifiable`** — and each accepts *both* registers' phrasing of the same condition: `all four gates pass` (comment, header) and `'promote-ready' verdict, dev-sync, commit-lint AND ship-tree all pass` (help) are one case in two voices, `a gate refuses` vs `'not promotable' a gate refused…`, `could NOT check` vs `cannot verify`. The comment also contains **three arrows that are prose, not cases** — its divergence-guard branches (`same commit + contradiction -> child_phrase():`, `different commits -> HEAD moved…`, `a commit we could not read -> the verdicts cannot be compared…`) — which a naive arrow-scan would have read as three extra verdicts: they are skipped because the right side does not **begin** with a verdict word, and the section asserts the result is exactly 3 rows per copy rather than merely equal ones.
- **Non-vacuity asserted, not assumed** (the lesson of (51)'s `P4`/`P5`): each copy must land on **exactly 3 entries carrying 3 distinct keys**; `spg_equal` **refuses two empty extractions and that refusal is itself asserted** alongside its acceptance of the three real equal maps; and **no copy may map any promote case to `error`** — the row's whole promise is *a refused promotion is not a system failure, the tool stays exit 0*, so `refused|error` in any register is red even when all three copies agree with each other (a comparison that only tests agreement would happily pass three identically wrong copies).
- **Six negative controls — a verdict flip and a key deletion per copy**, each asserting *first* that its plant changed the text it came from, so "caught" cannot be trivially true because the substitution matched nothing. The three **deletions** are caught by the map **shrinking 3 → 2**, not by an inequality that a lucky rewording could survive: renaming `--help`'s `'cannot verify'` label (that row's only key carrier — the rest of its text names neither condition), rewording the comment's `all four gates pass`, rewording the header's `everything we could NOT check`. **6/6 caught.**
- **Three live replays against the real files**, because a control over an extracted string only proves the extractor works: backup → `sed` plant → run → restore, each `md5`-verified back (`tools/system-status` and the suite both identical to their pre-plant digests). (a) plant in the tool's **comment** → `318 passed / 5 failed`; (b) plant in the tool's **`--help` block** → `318 / 5`; (c) plant in **this suite's own header** → `319 / 4`. In all three **every red is inside section 20** — the other 318/318/319 assertions stayed green because they pin the row's *behaviour*, which the plant does not change — which is the measurement that says the defect was real and localised. In each replay the affected copy's own N-control also went red: its "plant must differ from the text" guard refused to call a no-op *caught*, i.e. the control machinery behaved correctly under a plant it did not author.
- **Measured — the arithmetic closes**: `bash tests/test_system_status_promote_gates.sh` → **323 / 0 / 0** (was **295**; **+28** = 3 scoping + 9 per-copy (3 copies × 3) + 5 comparisons + 3 promise + 2 claim + 6 controls). `tools/regression-run` on the edited tree → **57 suites, 5022 passed, 2 failed**, i.e. **4996 + 28 = 5024** total assertions, and the 2 reds are `test_gladex_monitor.sh`'s dirty-tree pair **A3/A15** (`git status` sees this run's own uncommitted files — the pair `[0.4.107]` documents as red-before-commit/green-after). Gates read **after** the edits, none carried from before them: `tools/repo-lint --format json` → **exit 0, `ok true`, `failures []`, `duplicates []`, `citations_missing []`, `go_compile.ok true`** (it reads the **COMMITTED** `CHANGELOG.md`, so it still reports **116** entries and 3510 citations until this commit lands — re-read post-commit below the fold of this entry's history); `tools/queue-source-check --format json` → **exit 0, `violations []`, `newest_version 0.4.112`, pointer-only, 111 frozen item lines, `newest_queue_items 0`, `queue_headings_raw 70 == sections 69 + fenced 1`**; `test_repo_lint` **420 / 0**, `test_queue_source` **181 / 0**, `test_changelog_api` **86 / 0**, `test_changelog_mobile` **125 / 0**, `test_app_version` **39 / 0**, `test_cli_version` **35 / 0**; `tools/system-status --format json` → **rc 0, `overall ok`, `errors 0`**, with the same three standing `WARN` rows the investor's or the reviewer's to move (`cloud` Nextcloud `installed:false`, `SOA:gladex.de` placeholder MNAME, `promote-gates` stale `VERDICT-20260926-000744`).
- **`tools/system-status` is byte-unchanged** — both of its copies were already correct and already agreed; only the readers were missing, which is the whole point of (57)/(59). `tools/REGISTRY.md`'s promote-gates bullet carries the new count (**295 → 323**, mutations still **39** — the six new controls are labelled `20 N1…N6`, not `M`s) plus a section-20 description naming the three live replays.
- **Deliberately not done**: **no word-for-word comparison between registers** — `all four gates pass` and `'promote-ready' verdict… all pass` would be red on layout alone, so the declared granularity is (key → verdict) across registers, and the seam is **disclosed in the section's own comment**: a rewording that *keeps* its key (`a gate refuses` → `a gate said no`) stays green across registers — the compare-or-declare choice already parked as **(54)**/**(56)**. **No other suite extended** (`test_repo_lint.sh` 420 and `test_queue_source.sh` 181 byte-unchanged), **no rule added to any tool**, **no promote executed**, no service restarted, no certificate touched, **zero DNS writes**, no mail sent, Nextcloud untouched (`{"installed":false}`, §14 open), `/opt/cloud` untouched, Immich untouched (`{"res":"pong"}`, 8 accounts), and INBOX line 468 answered rather than pretended executed.
- **Staging hazard — the standing rule held**: `git status --porcelain` immediately before staging lists exactly this run's three files (`CHANGELOG.md`, `agent-logs/PROGRESS.md` is the fourth when written, `tests/test_system_status_promote_gates.sh`, `tools/REGISTRY.md`), staged **explicitly by path, never `git add -A`**, which has swept unfinished entries from other identities **eight** times in this file. Author resolves to `Atlas <atlas@gladex.de>`.
- **Next-candidate queued, not actioned**: carry items (2), (4)–(8), (14), (15), (17), (18), (20), (22), (23), (25), (28), (32), (36), (39), (42), (44), (47), (48), (52), (53), (54), (56) and (60) from the 07:2xZ entry unchanged — item **(59) struck — EXECUTED by `[0.4.112]`** (three prose copies of the promote-gates contract, none read by any assertion; section 20 added with two registers, three keys from promote's own vocabulary, six negative controls and three live replays, suite **295 → 323**, every plant's red confined to the new section). New from this run: **(61)** (57)'s per-child sweep stopped at the four children that *have* a `# Contract …` comment — `cloud`, `investor-duty` and `dns` also carry `--help` verdict blocks (L66/L85/L100) with no such comment, so the measurement never asked whether their contract text has a *second* copy somewhere else (`tools/REGISTRY.md`, a test header, a docs table) — worth a `grep` of each block's first verdict line across `tools/ REGISTRY.md/ tests/ docs/` before declaring those three copy-free, not actioned because (59) had to close the last *named* claim first. **(62)** the register parser's wrapped-condition rule is `indent <= 4`, so a condition wrapped across three lines with its **middle** line at indent 5 is read as detail and dropped — measured on this run's own header row 3, whose extracted condition reads `… (garbage output, no gates, dry_run not child)` with the middle line `true, a `ready` that contradicts its own gates, our timeout, a missing` silently skipped. It cannot cause a false green (the 3-key bound goes red if a key lived only there), but it makes the failure *look* like a dropped condition rather than a dropped line — worth either a rule comment in the parser or a key placed on such a line to see which, not actioned because the cost is a false red, not a false pass.
- **Safety (re-stated, nothing moved since)**: model spend **0.00** (`*-free` only), **`BUDGET.md` untouched (1.50 spent / 3.50 remaining**, month 2026-09, allowance 5.00), **zero DNS writes** (no `pdns-api.py` call), no paid API key configured, **no secret read or printed** (`/root/.pdns-token` and every `*.secret` untouched; no password, token or credential-shaped value in any prompt, log, file or commit — the only such fact recorded anywhere remains the investor's own declared rule "password = own email"), **no service restarted, no certificate touched, no promote executed, no mail sent**, Nextcloud untouched (`{"installed":false}`, §14 open), `/opt/cloud` untouched, Immich untouched (`{"res":"pong"}`, 8 accounts, library baseline unchanged), and both message DBs touched only by this run's STEP 0 reply (dev 105 / prod 71).
EOF
tail -2 agent-logs/PROGRESS.md | head -1 | cut -c1-80; wc -l agent-logs/PROGRESS.md
- **Next-candidate queued, not actioned**: carry items (2), (4)–(8), (14), (15
6111 agent-logs/PROGRESS.md
> build · mimo-v2.6-flash-free
$ cd /data/repo && ./tools/queue-source-check --format json | python3 -c 'import json,sys; d=json.load(sys.stdin); print(d["exit_code"], d["ok"], d["violations"], d["detail"])'; git status --porcelain
0 True [] one queue: [0.4.112] pointer-only, 111 item line(s) frozen across 69 section(s), 73 PROGRESS bullet(s)
 M CHANGELOG.md
 M agent-logs/PROGRESS.md
 M tests/test_system_status_promote_gates.sh
 M tools/REGISTRY.md
> build · mimo-v2.6-flash-free
$ cd /data/repo && git add CHANGELOG.md agent-logs/PROGRESS.md tests/test_system_status_promote_gates.sh tools/REGISTRY.md && git status --porcelain && git commit -q -m "test: [0.4.112] promote-gates contract had three prose copies with no reader — section 20 compares them, suite 295 -> 323

Queue item (59): tools/system-status's '# Contract (pinned by
tests/test_system_status_promote_gates.sh):' sentence had no reader, and
neither --help's promote-gates verdict block nor this suite's own header was
compared against anything. New section 20 reduces all three copies to one
(condition key -> verdict) map over three keys from promote's own vocabulary
(all-pass / refused / unverifiable - this contract has no exit codes), across
the two registers, with a 3-row/3-distinct-key bound per copy, the 'no promote
case is ever an error' promise, non-vacuity (empty maps refused and that
refusal asserted), the claim-has-reader check and six negative controls
(a verdict flip and a key deletion per copy, deletions caught by map 3 -> 2).

Verified by three live replays against the real files, each md5-restored:
planting the drift in the tool's comment -> 318/5, in its --help block ->
318/5, in this suite's own header -> 319/4, every red inside section 20.
Regression: 57 suites / 5024 assertions (4996 + 28), 2 reds = monitor's
dirty-tree pair, green after this commit. tools/system-status byte-unchanged.

STEP 0 answered first: 0 unread investor messages both DBs, fresh-probe reply
dev 105 / prod 71." && git log --oneline -1 && git status --porcelain
M  CHANGELOG.md
M  agent-logs/PROGRESS.md
M  tests/test_system_status_promote_gates.sh
M  tools/REGISTRY.md
3973fd6 test: [0.4.112] promote-gates contract had three prose copies with no reader — section 20 compares them, suite 295 -> 323
> build · mimo-v2.6-flash-free
$ cd /data/repo && ./tools/regression-run > /tmp/opencode/regr-post.log 2>&1; echo "regression rc=$?"; tail -3 /tmp/opencode/regr-post.log
Command moved to the background (shell ID: sh_0e7352a4e0014tus6W6eKcLYRa).
Output is streaming to: /root/.local/share/opencode/shell/fa3e85b2145a60c127b5c9261e76a7ba6d520299/sh_0e7352a4e0014tus6W6eKcLYRa.out
$ cd /data/repo && ./tools/repo-lint --format json | python3 -c 'import json,sys; d=json.load(sys.stdin); print("repo-lint", d["exit_code"], d["ok"], d["failures"], "|", d["changelog_version"]["summary"], "| go_compile", d["go_compile"]["ok"])'; ./tools/queue-source-check --format json | python3 -c 'import json,sys; d=json.load(sys.stdin); print("queue", d["exit_code"], d["ok"], d["violations"], d["rules"]["newest_version"])'
repo-lint 0 True [] | 117 changelog version heading(s), 117 unique, 3519 citation(s) checked, 0 missing | go_compile True
queue 0 True [] 0.4.112
> build · mimo-v2.6-flash-free
$ cd /data/repo && git push origin main 2>&1 | tail -3; git status -sb | head -2
To git://git.gladex.de/gladex.git
   2c4e750..3973fd6  main -> main
## main...origin/main
> build · mimo-v2.6-flash-free
$ cd /data/repo && md5sum CHANGELOG.md agent-logs/PROGRESS.md tests/test_system_status_promote_gates.sh tools/REGISTRY.md > /tmp/opencode/post.md5; for f in CHANGELOG.md agent-logs/PROGRESS.md tests/test_system_status_promote_gates.sh tools/REGISTRY.md; do a=$(md5sum "$f" | cut -d' ' -f1); b=$(git show HEAD:"$f" | md5sum | cut -d' ' -f1); [ "$a" = "$b" ] && echo "MATCH $f" || echo "DIFF $f"; done; git show HEAD:CHANGELOG.md | grep -c '^## \[0.4.112\]'; git show HEAD:tests/test_system_status_promote_gates.sh | grep -c '=== 20\.'
MATCH CHANGELOG.md
MATCH agent-logs/PROGRESS.md
MATCH tests/test_system_status_promote_gates.sh
MATCH tools/REGISTRY.md
1
1
> build · mimo-v2.6-flash-free
Committed `3973fd6` and pushed. Running the full post-commit regression now (57 suites, ~9 min); I'll confirm the results and re-check STEP 0 when it finishes.
$ sleep 240; tail -c 400 /tmp/opencode/regr-post.log 2>/dev/null || echo "no output yet"
(no output)
> build · mimo-v2.6-flash-free
Waiting on the regression run — it writes its report at the end. I'll continue as soon as it completes.
exit=0

Select another run log from the list above. Only files matching run-YYYYMMDD-HHMMSS-N.log are readable.

[Mon Sep 28 12:03:47 2026] 127.0.0.1:34440 Accepted
[Mon Sep 28 12:03:47 2026] 127.0.0.1:34440 Closing
[Mon Sep 28 12:03:49 2026] 127.0.0.1:34452 Accepted
[Mon Sep 28 12:03:49 2026] 127.0.0.1:34452 Closing
[Mon Sep 28 12:04:47 2026] 127.0.0.1:45952 Accepted
[Mon Sep 28 12:04:47 2026] 127.0.0.1:45952 Closing
[Mon Sep 28 12:05:47 2026] 127.0.0.1:34442 Accepted
[Mon Sep 28 12:05:47 2026] 127.0.0.1:34442 Closing
[Mon Sep 28 12:06:47 2026] 127.0.0.1:40214 Accepted
[Mon Sep 28 12:06:47 2026] 127.0.0.1:40214 Closing
[Mon Sep 28 12:07:47 2026] 127.0.0.1:39194 Accepted
[Mon Sep 28 12:07:47 2026] 127.0.0.1:39194 Closing
[Mon Sep 28 12:08:47 2026] 127.0.0.1:33522 Accepted
[Mon Sep 28 12:08:47 2026] 127.0.0.1:33522 Closing
[Mon Sep 28 12:09:47 2026] 127.0.0.1:51844 Accepted
[Mon Sep 28 12:09:47 2026] 127.0.0.1:51844 Closing
[Mon Sep 28 12:10:47 2026] 127.0.0.1:40130 Accepted
[Mon Sep 28 12:10:47 2026] 127.0.0.1:40130 Closing
[Mon Sep 28 12:11:47 2026] 127.0.0.1:56502 Accepted
[Mon Sep 28 12:11:47 2026] 127.0.0.1:56502 Closing
[Mon Sep 28 12:12:47 2026] 127.0.0.1:54546 Accepted
[Mon Sep 28 12:12:47 2026] 127.0.0.1:54546 Closing
[Mon Sep 28 12:13:47 2026] 127.0.0.1:48154 Accepted
[Mon Sep 28 12:13:47 2026] 127.0.0.1:48154 Closing
[Mon Sep 28 12:14:47 2026] 127.0.0.1:34470 Accepted
[Mon Sep 28 12:14:47 2026] 127.0.0.1:34470 Closing
[Mon Sep 28 12:28:41 2026] 127.0.0.1:44172 Accepted
[Mon Sep 28 12:28:41 2026] 127.0.0.1:44172 Closing
[Mon Sep 28 12:28:42 2026] 127.0.0.1:44182 Accepted
[Mon Sep 28 12:28:42 2026] 127.0.0.1:44182 Closing
[Mon Sep 28 12:30:54 2026] 127.0.0.1:59176 Accepted
[Mon Sep 28 12:30:54 2026] 127.0.0.1:59176 Closing
[Mon Sep 28 12:31:54 2026] 127.0.0.1:41156 Accepted
[Mon Sep 28 12:31:54 2026] 127.0.0.1:41156 Closing
[Mon Sep 28 12:32:54 2026] 127.0.0.1:60436 Accepted
[Mon Sep 28 12:32:54 2026] 127.0.0.1:60436 Closing
[Mon Sep 28 12:33:54 2026] 127.0.0.1:47408 Accepted
[Mon Sep 28 12:33:54 2026] 127.0.0.1:47408 Closing
[Mon Sep 28 12:34:54 2026] 127.0.0.1:49594 Accepted
[Mon Sep 28 12:34:54 2026] 127.0.0.1:49594 Closing
[Mon Sep 28 12:35:54 2026] 127.0.0.1:38868 Accepted
[Mon Sep 28 12:35:54 2026] 127.0.0.1:38868 Closing
[Mon Sep 28 12:36:54 2026] 127.0.0.1:60072 Accepted
[Mon Sep 28 12:36:54 2026] 127.0.0.1:60072 Closing
[Mon Sep 28 12:37:54 2026] 127.0.0.1:53554 Accepted
[Mon Sep 28 12:37:54 2026] 127.0.0.1:53554 Closing
[Mon Sep 28 12:38:54 2026] 127.0.0.1:45994 Accepted
[Mon Sep 28 12:38:54 2026] 127.0.0.1:45994 Closing
[Mon Sep 28 12:39:54 2026] 127.0.0.1:53700 Accepted
[Mon Sep 28 12:39:54 2026] 127.0.0.1:53700 Closing
[Mon Sep 28 12:40:54 2026] 127.0.0.1:54316 Accepted
[Mon Sep 28 12:40:54 2026] 127.0.0.1:54316 Closing
[Mon Sep 28 12:41:42 2026] 127.0.0.1:38460 Accepted
[Mon Sep 28 12:41:42 2026] 127.0.0.1:38460 Closing
[Mon Sep 28 12:41:42 2026] 127.0.0.1:38470 Accepted
[Mon Sep 28 12:41:42 2026] 127.0.0.1:38470 Closing
[Mon Sep 28 12:41:43 2026] 127.0.0.1:38478 Accepted
[Mon Sep 28 12:41:43 2026] 127.0.0.1:38478 Closing
[Mon Sep 28 12:41:46 2026] 127.0.0.1:59848 Accepted