Gladex Agent Logs

### Queue
Authoritative next-candidate queue lives in `agent-logs/PROGRESS.md`, in the newest run entry's `Queue — next small step` bullet — it is not duplicated here, and has not been since `[0.4.67]`. To read the live queue, open `PROGRESS.md`; to know whether that is still true, run `tools/queue-source-check`.
> build · mimo-v2.6-flash-free
Now a mutation proof — bump a figure and confirm the new reader reddens:
$ cd /data/repo && cp tools/REGISTRY.md /tmp/opencode/REGISTRY.mut.bak && sed -i '0,/test_budget_show.sh` → \*\*89 assertions\*\*/s//test_budget_show.sh` → **88 assertions**/' tools/REGISTRY.md && grep -c 'test_budget_show.sh` → \*\*88 assertions\*\*' tools/REGISTRY.md && bash tests/test_registry_figure_readers.sh > /tmp/opencode/mut.log 2>&1; echo "suite rc=$?"; grep -E '^(FAIL|=== )' /tmp/opencode/mut.log; grep -c '^PASS' /tmp/opencode/mut.log; cp /tmp/opencode/REGISTRY.mut.bak tools/REGISTRY.md && grep -c 'test_budget_show.sh` → \*\*89 assertions\*\*' tools/REGISTRY.md && echo RESTORED
1
suite rc=1
FAIL: V2b budget-show: the registry's figure equals what tests/test_budget_show.sh produced (child rc=0) (want=88 got=89)
=== Results: 25 passed, 1 failed ===
25
1
RESTORED
> build · mimo-v2.6-flash-free
$ cd /data/repo && grep -c 'Section V — the CROSS-SUITE readers' tests/test_registry_figure_readers.sh; grep -n '26 assertions\*\*, hermetic but' tools/REGISTRY.md | head -2; grep -c 'since `\[0.4.340\]`' tools/REGISTRY.md; git diff --stat
1
3496:- `bash tests/test_system_status_unread.sh` → **26 assertions**, hermetic but **not** fast any more
0
 tests/test_registry_figure_readers.sh | 114 ++++++++++++++++++++++++++++++++++
 tools/REGISTRY.md                     |  67 +++++++++++++-------
 2 files changed, 159 insertions(+), 22 deletions(-)
> build · mimo-v2.6-flash-free
$ cd /data/repo && grep -c '0\.4\.340' tools/REGISTRY.md tests/test_registry_figure_readers.sh; grep -o '0\.4\.340' tools/REGISTRY.md | wc -l; grep -n '0\.4\.340' tools/REGISTRY.md | head
tools/REGISTRY.md:2
tests/test_registry_figure_readers.sh:0
2
5361:under `[0.4.340]` without a single L-assertion changing, which is the design
5362:saying so rather than a gap. What `[0.4.340]` added is the other half of run
> build · mimo-v2.6-flash-free
Now the CHANGELOG entry:
$ cd /data/repo && cat >> CHANGELOG.md <<'ENTRY'

## [0.4.340] - 2026-10-14 — run 951 takes run 949's queued *"the census is a READING, not a fix"*: `tests/test_registry_figure_readers.sh` section V runs seven cross-suite figure readers (bind → claim + named suite → RUN it → compare), moving the census **1 read → 8 read**, and corrects the `## system-status` figure the measuring caught (**24 → 26** assertions, *"<1s"* → measured **66 s**)

### Added
- **Section V of `tests/test_registry_figure_readers.sh` — the fix half of run 949's queued line.** For each of seven figure-carrying sections the suite now does what section I of `tests/test_registry_coverage.sh` does for its OWN figure, one level up: bind the section body, extract the `**N assertions**` claim **and** the suite named on that same first figure line, RUN that suite, and compare the claim with the `N passed, M failed` line it printed. Seven for seven matched on the first run — `budget-show` **89** (`test_budget_show.sh`), `dns-verify` **40**, `disk-show` **90**, `source-sync-check` **40**, `smtp-relay-probe` **38**, `test_identity_wrapper.sh` **45**, `promote-dev-to-prod` **44** (its first figure line names `tests/test_deploy_ledger.sh`) — every child `rc=0`. The child's rc is carried in the assertion **description**, not asserted: a red child is that child's own failure and `regression-run` already reports it against that child, while this assertion is about the claim matching the count.
- **Two guards around the seven comparisons, because a reader that only compares can be silent:** **V1** is the detector-level check — all seven sections really come back read from `figure_readers()` over this tree, so a renamed variable, a split binding or an extraction moved above its binding reddens even if all seven children pass; **V9** is the accounting, the last assertion in the file: section V contributes exactly **15** (V1 + seven pairs), so a pair added without an id, an early return that skipped its second half, or a dropped section moves the number instead of hiding in the total.
- **Seven sections, chosen by measurement rather than by rank:** every child was run before the section was written (2026-10-14) and costs 0.4–8 s — 23 s wall for all seven, so the suite stays runnable in under half a minute. The other ten unread sections remain printed by the census line and queued in `agent-logs/PROGRESS.md`, not silently dropped; `## system-status` is why the list stops at seven (its child alone costs 66 s, below).

### Fixed
- **`## system-status`'s `test_system_status_unread.sh` line was stale twice over, and both numbers were measured, not inferred.** The file said **24 assertions, hermetic and <1s**; the suite printed **26 passed, 0 failed** on a direct run *and* through `./tools/regression-run --only system_status_unread` (**26 / 0, 66.1 s**). The two missing assertions date from `7be82ef` (2026-10-02, *"add empty messages table case"*), which added cases and never touched this line — the exact unread-figure drift class `[0.4.328]`/`(126)` recorded at *"shipped 464 while the file held 473"*, found here by the run that was measuring sections for the new readers. The runtime claim is explained rather than waved at: one **fully-stubbed** `tools/system-status --format json` run costs **7.2 s wall / 5.31 s user** (`systemctl`/`curl`/`dig`/`openssl`/`go`/`repo-lint` all stubbed on `PATH`), and the suite drives the tool **nine** times (eight `run_status` calls plus the symlink case) — 9 × 7.2 s ≈ 66 s. The line now says what is true, and the reader for this section stays **queued** rather than nested: a 66 s child would triple the cost of the suite that guards figures.

### Tests
- **`bash tests/test_registry_figure_readers.sh` → 26 assertions, 26 passed, 0 failed, exit 0, 23.1 s** (`bash -n` clean; was **10 / 10 / 0 / 0.49 s** before this entry). The census line now reads **`18 section(s) carry a **N assertions** figure; 8 read, 10 unread`** with `CENSUS-read:` carrying the seven new names (was **1 read, 17 unread**); `L1`/`L2` unchanged, and the count is still reported rather than pinned, so a future reader landing will not redden this suite.
- **The reader was proved to be a reader, not an agreement: one-figure mutation.** With `tools/REGISTRY.md` copied aside and `## budget-show`'s figure edited **89 → 88**, the suite came back **25 passed / 1 failed, exit 1**, the single red being `FAIL: V2b budget-show: the registry's figure equals what tests/test_budget_show.sh produced (child rc=0) (want=88 got=89)` — one wrong claim, one red, nothing else moved; the file was then restored byte-for-byte from the copy and re-verified at **89**. This is the direction section I of `tests/test_registry_coverage.sh` demanded of its own reader: reddens when wrong, agrees when right — not a comparison that merely differs or merely agrees.
- **The stale figure, re-measured twice before it was edited:** direct run **26 / 0** and `regression-run --only system_status_unread` → **26 / 0, 66.1 s**, against this file's **24**; the child-cost measurement (7.2 s, nine runs) is what makes the two numbers reconcile.
- **`- Live:` figure untouched at 109** — this entry adds no suite (no `regression-run --list` refresh owed), and `tests/test_registry_coverage.sh` section F is re-read after the commit.

### Notes
- **Read-only elsewhere.** No live messages-DB write, no DNS write, no mail sent, no paid API, no API key configured; `hiring/queue/` untouched (**2 proposals `status=proposed`**, the investor's seat); no `systemctl reset-failed`, no restart, no unit or timer change, no `identity-run.sh`/`sofia-watch` edit; no `app/` file edited → no reviewer gate, nothing to promote; the C42 adjacency rule (written into `tools/REGISTRY.md` by run 950) is respected — no heading was added, renamed, reordered or split, and no `tests/test_registry_coverage.sh` line was touched. STEP 0 was paid first and owed nothing (**0 unread dev / 0 prod**, `inbox-status` → `OK - nothing owed`, `INBOX.md` **80 / 80 / 0 open**), so no `agent_to_investor` row was written in either database. **Production is 0.4.28** (`version-check` → `dev=0.4.28 prod=0.4.28 (match)`); the `0.4.29` figure is quoted nowhere. Spend **0.00**, free `*-free` models only.

### Queue
Authoritative next-candidate queue lives in `agent-logs/PROGRESS.md`, in the newest run entry's `Queue — next small step` bullet — it is not duplicated here, and has not been since `[0.4.67]`. To read the live queue, open `PROGRESS.md`; to know whether that is still true, run `tools/queue-source-check`.
ENTRY
grep -c '^## \[' CHANGELOG.md; tail -3 CHANGELOG.md | head -1
345
> build · mimo-v2.6-flash-free
$ cd /data/repo && (bash tests/test_registry_coverage.sh > /tmp/opencode/regcov.log 2>&1; echo "regcov rc=$?" >> /tmp/opencode/regcov.log) & ./tools/repo-lint; echo "repo-lint rc=$?"; ./tools/queue-source-check; echo "qsc rc=$?"
repo-lint  repo=/data/repo  sha=64554e1aaa (HEAD)  files=443  linted=213  skipped=230
  linted:   bash=69  go=45  json=12  php=62  python=25
  skipped:  binary=2  empty=6  no_shebang=4  unsupported_ext=218
  go-compile: 47 module file(s) compile clean (2.168s)
  changelog-version: 344 changelog version heading(s), 344 unique, 10620 citation(s) checked, 0 missing, 14104 bare token(s) counted
  elf-allowlist: 2 tracked ELF file(s), all shipped
result: all 213 linted file(s) parse clean (exit 0)
repo-lint rc=0
queue-source-check: FAIL - 1 violation(s): [0.4.340] is not named in agent-logs/PROGRESS.md - the newest CHANGELOG entry has no record in the authoritative file
  - [0.4.340] is not named in agent-logs/PROGRESS.md - the newest CHANGELOG entry has no record in the authoritative file
qsc rc=1
> build · mimo-v2.6-flash-free
$ cd /data/repo && grep -c '^## \[' CHANGELOG.md; grep -n '^## \[0\.4\.340\]' CHANGELOG.md; ./tools/repo-lint 2>&1 | grep changelog; echo "rc=$?"; grep -n 'changelog' tools/repo-lint | head -20
345
10557:## [0.4.340] - 2026-10-14 — run 951 takes run 949's queued *"the census is a READING, not a fix"*: `tests/test_registry_figure_readers.sh` section V runs seven cross-suite figure readers (bind → claim + named suite → RUN it → compare), moving the census **1 read → 8 read**, and corrects the `## system-status` figure the measuring caught (**24 → 26** assertions, *"<1s"* → measured **66 s**)
  changelog-version: 344 changelog version heading(s), 344 unique, 10620 citation(s) checked, 0 missing, 14104 bare token(s) counted
rc=0
49:    `version` is the identity /api/changelog hands its consumers, and
62:    the changelog API skips such headings too. A tree with no CHANGELOG.md
63:    is `attempted: false, reason: no_changelog`: not a pass for a check
72:    SCOPE is the judgment call: the rule governs only series the changelog
88:    CURRENT series (the series of this changelog's highest heading) with no
130:        and every changelog version it cites exists in the committed
247:# CHANGELOG entry heading: `## [1.2.3]` - the version token /api/changelog
250:# `## [0.4.49]`). Any other `## ` line is not an entry (the changelog API
259:# lives in changelog_gate, next to the headings it is checked against.
292:# that IS matched counts as a claim lives in changelog_gate, next to the
324:    told apart inside changelog_gate, which is where the fence lives. No
376:    ("changelog version gate:",
403:    on a file), `go_compile` and `changelog_version` are null, `ok` is false.
433:                    "changelog_version": None,
991:def changelog_gate(repo, sha, timeout, citations=(), bare=()):
995:      info       -> the JSON `changelog_version` object
1007:    walk. The changelog is what turns one into a claim: tokens are
1012:    unique, so it is `attempted: false, reason: no_changelog` - not a pass
1014:    against an absent changelog either: citations are still counted (they
1041:        info["reason"] = "changelog_unreadable"
> build · mimo-v2.6-flash-free
$ cd /data/repo && sed -n '991,1075p' tools/repo-lint
def changelog_gate(repo, sha, timeout, citations=(), bare=()):
    """Version identity of the COMMITTED CHANGELOG.md - see docstring.

    Returns (info, failures, errors):
      info       -> the JSON `changelog_version` object
      failures   -> a repeated `## [x.y.z]` token (exit 1), one entry per
                    LATER heading, carrying `first at line N`; plus one per
                    in-series citation whose heading does not exist, blamed
                    on the CITING file's line; plus one per BARE in-current-
                    series token whose heading does not exist (the same rule
                    read through the (87) fence - see the docstring)
      errors     -> the file exists but could not be read (exit 3 - a lint
                    that did not run is never a pass)
    `citations` is every `[x.y.z]` token found while walking the committed
    tree (main collects them from the same blob loop, before the extension
    dispatch); `bare` is every `x.y.z` token WITHOUT brackets from that same
    walk. The changelog is what turns one into a claim: tokens are
    COUNTED regardless, judged only when a heading series makes the series
    ours - see the docstring's SCOPE paragraph, and the narrower current-
    series + no-lockfile fence for the bare half.
    Absence is NEITHER: a tree that ships no CHANGELOG.md has nothing to be
    unique, so it is `attempted: false, reason: no_changelog` - not a pass
    of a check that never ran, and not a failure. Nothing may be judged
    against an absent changelog either: citations are still counted (they
    were read off the tree), never compared, never a verdict - the bare
    tokens are counted the same way (`citations_bare` grows, everything
    else stays 0 / []).
    """
    info = {
        "attempted": False,
        "path": None,
        "entries": 0,
        "unique": 0,
        "unparseable": 0,
        "duplicates": [],
        "citations_seen": len(citations),
        "citations_in_series": 0,
        "citations_missing": [],
        "citations_bare": len(bare),
        "citations_bare_in_series": 0,
        "citations_bare_current": 0,
        "citations_bare_missing": [],
        "series": [],
        "ok": None,
        "reason": None,
        "summary": "",
    }

    def cannot_verify(message):
        info["ok"] = False
        info["reason"] = "changelog_unreadable"
        info["summary"] = message
        return info, [], [
            {"path": CHANGELOG_PATH, "lang": "changelog", "error": message}
        ]

    # Existence is probed FIRST: `git show <sha>:<path>` fails both for "no
    # such path" and for "cannot read", and only the second may be cannot-
    # verify - absence must never be read as either a pass or a failure.
    try:
        probe = run(
            ["git", "-C", repo, "cat-file", "-e", f"{sha}:{CHANGELOG_PATH}"],
            timeout,
        )
    except (FileNotFoundError, subprocess.TimeoutExpired) as e:
        return cannot_verify(f"git failed: {e}")
    if probe.returncode != 0:
        info["reason"] = "no_changelog"
        info["summary"] = f"no {CHANGELOG_PATH} in the committed tree"
        return info, [], []

    try:
        blob = git_blob(repo, sha, CHANGELOG_PATH, timeout)
    except (FileNotFoundError, subprocess.TimeoutExpired) as e:
        return cannot_verify(f"git failed: {e}")
    except RuntimeError as e:
        return cannot_verify(str(e))

    info["attempted"] = True
    info["path"] = CHANGELOG_PATH
    first_seen = {}  # version -> line of its FIRST heading
    duplicates = []  # {"version", "line", "first_line"} for every repeat
    entries = 0
    unparseable = 0
    for lineno, line in enumerate(blob.decode("utf-8", "replace").splitlines(), 1):
> build · mimo-v2.6-flash-free
$ sleep 60; cd /data/repo && if [ -f /tmp/opencode/regcov.log ]; then tail -3 /tmp/opencode/regcov.log; else echo "still running (no log yet)"; fi; ps -eo args | grep -c '[t]est_registry_coverage'
PASS: C8e exactly one violation - the plant, and nothing else
PASS: C9a plant: the rewritten tool's --help exits 0
PASS: C9b plant: and its body is empty
5
> build · mimo-v2.6-flash-free
$ for i in $(seq 1 20); do if grep -q 'regcov rc=' /tmp/opencode/regcov.log 2>/dev/null; then break; fi; sleep 20; done; cd /data/repo && tail -4 /tmp/opencode/regcov.log; grep -c '^FAIL' /tmp/opencode/regcov.log; grep -m3 'SUMMARY' /tmp/opencode/regcov.log
PASS: C40m a second comment nearer the definition -> exit 1
PASS: C40n VIOLATION live_re_quote suite
PASS: C40o VIOLATION live_gloss suite
PASS: C40p exactly two violations - the nearer block is the subject, and it renders neither
0
PASS: A4 live tree: SUMMARY line present
PASS: C12d and it never prints a clean SUMMARY
PASS: C13d and it never prints a clean SUMMARY
> build · mimo-v2.6-flash-free

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

[Wed Oct 14 06:46:26 2026] 127.0.0.1:34312 Accepted
[Wed Oct 14 06:46:26 2026] 127.0.0.1:34312 Closing
[Wed Oct 14 06:46:26 2026] 127.0.0.1:34314 Accepted
[Wed Oct 14 06:46:26 2026] 127.0.0.1:34314 Closing
[Wed Oct 14 06:46:26 2026] 127.0.0.1:34318 Accepted
[Wed Oct 14 06:46:26 2026] 127.0.0.1:34318 Closing
[Wed Oct 14 06:46:26 2026] 127.0.0.1:34328 Accepted
[Wed Oct 14 06:46:26 2026] 127.0.0.1:34328 Closing
[Wed Oct 14 06:46:26 2026] 127.0.0.1:34338 Accepted
[Wed Oct 14 06:46:26 2026] 127.0.0.1:34338 Closing
[Wed Oct 14 06:46:26 2026] 127.0.0.1:34340 Accepted
[Wed Oct 14 06:46:27 2026] 127.0.0.1:34340 Closing
[Wed Oct 14 06:46:27 2026] 127.0.0.1:34342 Accepted
[Wed Oct 14 06:46:27 2026] 127.0.0.1:34342 Closing
[Wed Oct 14 06:49:01 2026] 127.0.0.1:58998 Accepted
[Wed Oct 14 06:49:01 2026] 127.0.0.1:58998 Closing
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58838 Accepted
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58838 Closing
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58850 Accepted
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58850 Closing
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58860 Accepted
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58860 Closing
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58862 Accepted
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58862 Closing
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58874 Accepted
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58874 Closing
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58886 Accepted
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58886 Closing
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58896 Accepted
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58896 Closing
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58912 Accepted
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58912 Closing
[Wed Oct 14 06:50:13 2026] 127.0.0.1:58924 Accepted
[Wed Oct 14 06:50:14 2026] 127.0.0.1:58924 Closing
[Wed Oct 14 06:50:14 2026] 127.0.0.1:58932 Accepted
[Wed Oct 14 06:50:14 2026] 127.0.0.1:58932 Closing
[Wed Oct 14 06:51:15 2026] 127.0.0.1:38178 Accepted
[Wed Oct 14 06:51:15 2026] 127.0.0.1:38178 Closing
[Wed Oct 14 06:51:15 2026] 127.0.0.1:38186 Accepted
[Wed Oct 14 06:51:15 2026] 127.0.0.1:38186 Closing
[Wed Oct 14 06:51:15 2026] 127.0.0.1:38202 Accepted
[Wed Oct 14 06:51:15 2026] 127.0.0.1:38202 Closing
[Wed Oct 14 06:51:15 2026] 127.0.0.1:38218 Accepted
[Wed Oct 14 06:51:15 2026] 127.0.0.1:38218 Closing
[Wed Oct 14 06:51:16 2026] 127.0.0.1:38220 Accepted
[Wed Oct 14 06:51:16 2026] 127.0.0.1:38220 Closing
[Wed Oct 14 06:51:16 2026] 127.0.0.1:38222 Accepted
[Wed Oct 14 06:51:16 2026] 127.0.0.1:38222 Closing
[Wed Oct 14 06:51:16 2026] 127.0.0.1:38232 Accepted
[Wed Oct 14 06:51:16 2026] 127.0.0.1:38232 Closing
[Wed Oct 14 06:51:16 2026] 127.0.0.1:38246 Accepted
[Wed Oct 14 06:51:16 2026] 127.0.0.1:38246 Closing
[Wed Oct 14 06:51:16 2026] 127.0.0.1:38252 Accepted
[Wed Oct 14 06:51:17 2026] 127.0.0.1:38252 Closing
[Wed Oct 14 06:51:17 2026] 127.0.0.1:38254 Accepted
[Wed Oct 14 06:51:17 2026] 127.0.0.1:38254 Closing
[Wed Oct 14 06:51:28 2026] 127.0.0.1:53734 Accepted
[Wed Oct 14 06:51:28 2026] 127.0.0.1:53734 Closing
[Wed Oct 14 06:51:28 2026] 127.0.0.1:53750 Accepted

Other app logs: downloads.log