Gladex Agent Logs

 **Purpose**: Grade the **STEP 0 investor duty** — *"reply to ALL unread investor

> build · mimo-v2.6-flash-free
$ cd /data/repo && grep -rn "test_registry_coverage" --include="*.sh" --include="*.md" --include="*.php" . | grep -v "^./tests/test_registry_coverage.sh" | grep -v "^./agent-logs/" | head -20; echo "=== 67 assertions refs ==="; grep -rn "67 assertions\|67/67" --include="*.md" --include="*.sh" . | grep -v agent-logs | head
./CHANGELOG.md:4130:- **New `tests/test_registry_coverage.sh` (67 assertions, suite 57 → 58).** Section **A** runs the checker over the live tree and requires exit 0 with zero violations; **B** cross-checks every count against an *independent recount* of both inputs (tools, sections, tool sections, `--help` probed, `--help` answered, plus two floors), so a checker that silently stopped enumerating cannot pass by agreeing with itself; **C** is 13 negative controls; **D** pins this suite's own `--help`/exit-2 surface — the rule it enforces for everyone else.
./CHANGELOG.md:4139:- `bash tests/test_registry_coverage.sh` → **67 passed / 0 failed** (~11s, no network), both standalone and inside the full run.
./tools/REGISTRY.md:2882:## test_registry_coverage.sh
./tools/REGISTRY.md:2898:**Location**: `/data/repo/tests/test_registry_coverage.sh`
./tools/REGISTRY.md:2902:bash tests/test_registry_coverage.sh [--help]
./tools/REGISTRY.md:2935:- `bash tests/test_registry_coverage.sh` → **86 assertions**, ~24s, no network:
=== 67 assertions refs ===
./CHANGELOG.md:945:- **`tests/test_promote_gate.sh` (67 assertions) had to grow the same stub** as the three `system-status` suites did in `[0.4.39]`: its sandboxes model a repo whose child exists and reports a clean HEAD, because that suite pins only the verdict/dev-sync gates. Without it 13 of its assertions went red on the *new* gate — the right failure, in the right direction, and the reason the stub is commented in place.
./CHANGELOG.md:1765:- `repo-lint --format json` → exit 0, **171 files pre-append → 173 on the committed blob** (the two mailbox files of this run were still untracked when the number was drafted), 45 Go module files compile clean, `changelog_version` **67/67 pre-append → `entries 68, unique 68, duplicates []` on the post-commit re-read**.
./CHANGELOG.md:3910:  the 167 assertions that pinned them are unchanged.
./CHANGELOG.md:4130:- **New `tests/test_registry_coverage.sh` (67 assertions, suite 57 → 58).** Section **A** runs the checker over the live tree and requires exit 0 with zero violations; **B** cross-checks every count against an *independent recount* of both inputs (tools, sections, tool sections, `--help` probed, `--help` answered, plus two floors), so a checker that silently stopped enumerating cannot pass by agreeing with itself; **C** is 13 negative controls; **D** pins this suite's own `--help`/exit-2 surface — the rule it enforces for everyone else.
./tools/REGISTRY.md:2608:**Status**: Green ✅ 2026-09-27 (**348/348**; full regression **52 suites / 4267 assertions / 0 failed / 0 skipped, exit 0** — 29 shell = 2892, 23 PHP = 1375. Closure is arithmetic: the previous green run was **52 / 4243** with this suite at **324**, and this step adds **+24 (324 → 348) touching no other suite**, so 4243 + 24 = 4267 with 52 = 52 suites, and the shell split moves 2868 → 2892 by the same 24 while PHP stays 1375. The regression clause that follows was **carried unchanged at 41 / 3503 since `[0.4.67]`** — the tree had already reached 52 / 4210 by `[0.4.80]` without this line being touched, the same stale-line class the entry above records — so it is **refreshed here rather than silently extended**. History: 41 suites / 3503 / 0 failed / 21 shell = 2312, 20 PHP = 1191 at 291/291 on 2026-09-26. Was 40 / 3396 / 20 shell = 2205 with `[0.4.66]`; **`[0.4.67]` adds `tests/test_queue_source.sh` (+1 suite, +107 shell), so 2205 + 107 = 2312 and 3396 + 107 = 3503**, PHP untouched at 1191. Before that, `[0.4.66]` grew `tests/test_regression_run.sh` (+0 suites, +83 shell: 110 → 193), so 2122 + 83 = 2205 and 3313 + 83 = 3396, and `[0.4.65]` added that suite (+1 suite, +110 shell) — 2012 + 110 = 2122, 3203 + 110 = 3313. This line was itself two runs stale at `[0.4.64]` — `[0.4.63]` added `test_app_contrast.php` (+1, +185) without updating it, so 958 + 185 + 48 = 1191 and 2970 + 185 + 48 = 3203 came out then)
./STRUCTURE.md:265:   (67 assertions, 5 mutations — incl. the stdout line-buffering/output-order
> build · mimo-v2.6-flash-free
$ cd /data/repo && sed -n '260,268p' STRUCTURE.md; echo "=== who asserts REGISTRY suite counts? ==="; grep -rn "registry_coverage\|REGISTRY.md" tests/*.sh tests/*.php 2>/dev/null | grep -v test_registry_coverage | head
the env copy of `examples/workflows` shadows the repo on `/templates`). One
   `PROMOTED_TREES` list drives both the rsync loop and this freshness check, so they
   cannot drift, and a test asserts it matches `source-sync-check`'s tree set. Newest
   verdict wins, so a later REJECT overrides an earlier APPROVE; unparseable verdicts
   fail closed. `--force` bypasses loudly. Suite: `tests/test_promote_gate.sh`
   (67 assertions, 5 mutations — incl. the stdout line-buffering/output-order
   contract: stdout is line-buffered so a refusal can never overprint the
   earlier-emitted "Starting promotion" line in `2>&1` captures).
   **5b. Dev-sync gate (step 0b, same run)**: that tool deploys **repo → prod**, not
=== who asserts REGISTRY suite counts? ===
tests/test_domain_availability.sh:33:REGISTRY="$REPO/tools/REGISTRY.md"
tests/test_ip_drift_check.sh:45:REGISTRY="$REPO/tools/REGISTRY.md"
tests/test_ip_drift_check.sh:479:    bad "REGISTRY.md not found"
tests/test_ip_drift_cron.sh:40:REGISTRY="$REPO/tools/REGISTRY.md"
tests/test_ip_drift_cron.sh:319:    bad "REGISTRY.md not found"
tests/test_pdns_api.sh:39:REGISTRY="$REPO/tools/REGISTRY.md"
tests/test_pdns_api.sh:620:    bad "REGISTRY.md not found"
tests/test_queue_source.sh:614:# The module docstring, tools/REGISTRY.md's rules table and this suite all used
tests/test_queue_source.sh:749:REG_FILE="$(dirname "$TOOL")/REGISTRY.md"
tests/test_system_status_go_compile.sh:574:# tools/REGISTRY.md's 8-row `child evidence | status | detail` table is
> build · mimo-v2.6-flash-free
Now the docs. First the CHANGELOG entry `[0.4.114]`:
$ cd /data/repo && git status --porcelain && echo "--- tail 3 lines of CHANGELOG ---" && tail -3 CHANGELOG.md
M tests/test_registry_coverage.sh
 M tools/REGISTRY.md
--- tail 3 lines of CHANGELOG ---

### 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`.
> build · mimo-v2.6-flash-free
$ cd /data/repo && cat >> CHANGELOG.md <<'EOF'

## [0.4.114] - 2026-09-28 — queue item (64): a documented exit code is now PROBED, not merely registered — `tests/test_registry_coverage.sh` gains section E plus controls C14/C15, suite 67 → 86

### Why

`[0.4.113]` queued (64) as *"the `--help` claims exit 2 for unknown arguments and D2 measures it, but the tools side of the same claim (every registered tool's documented exit-2 line vs. its actual behaviour) is unchecked — `run_check` proves `--help` exits 0, not that the table under it is true"*. The fourth BOOTSTRAP field, `Exit codes`, was being asserted for **presence** only: a `tools/REGISTRY.md` section could promise `2 - Invalid arguments` while its tool returned `0` for the same argument, and every suite in the tree would stay green — the same shape as the unloved `pinned by this suite` comments (55)/(57)/(59), one child further along, this time inside the suite that was written to close that class.

Measured before writing anything, so the step knows what it is: **20/20** tools answer exit **2** on an unknown flag, fast (validation before any I/O — the whole sweep runs in ~2s), and **20/20** tables list a code 2, including the two shapes a naive parser misses — `immich-roundtrip`'s **wrapped header** (`**Exit codes** (precedence …` over two lines) and `smtp-relay-probe`'s **non-sequential list** (`1 4 3 0 2`, first-match-wins). The invariant already held, so this entry is a **guard, not a repair**: no tool changed, `tools/system-status` is byte-unchanged.

### Changed

- **`run_check()` probes each tool twice in the same branch as `--help`**: `timeout 5 <tool> --regcov-not-a-real-flag` must exit 2 (`VIOLATION badarg_contract <tool>:rc` otherwise), and the section's `**Exit codes**` block must really list a code 2 (`VIOLATION badarg_undocumented <tool>`). Both live in the one code path every control re-execs, so C14/C15 cannot be caught by a separately-written checker. The bound is **5s, not the `--help` probe's 20s**, and that is the assertion rather than a caveat: a tool that needs the network — or a hang — to discover that an argument is nonsense has failed the very rule its own table claims (`dns-verify`'s row spells it out: *"all validation runs before any network or DNS I/O"*).
- **A section with no `**Exit codes**` heading at all stays C7's `incomplete_section`** and is not re-reported by the new check — one plant, one finding — so the new reader checks the table's *content* where the old field check already owns its *presence*.
- **Section E (7 assertions)** reads A's own run: `badarg_run == help_run`, `badarg_ok == badarg_run`, `ec_run == help_run`, `ec_ok == ec_run`, a floor of **10** probes, plus a no-`badarg_*`-violation line. The floor and the two equalities are the non-vacuity half: a checker that stopped enumerating would report `0 == 0`.
- **Two new negative controls, one per copy of the claim.** **C14** models a tool with no argument validation at all (exit 0 on nonsense, `--help` still correct) → `badarg_contract version-check:rc`, exactly one violation. **C15** removes `budget-show`'s `` - `2` `` bullet from its table while the tool still exits 2 → `badarg_undocumented budget-show`, exactly one violation, with the tool's own behaviour asserted unchanged first so the plant cannot be hiding behind a second defect.
- **C8/C9's stubs were scoped to their own claim**: both now answer the bad-argument probe correctly (exit 2 off `--help`), because a stub failing *every* argv is caught twice and the suite's `exactly one violation` assertion would then be measuring the plant's sloppiness rather than the checker's precision.
- **`LIVE_OUT`**: section E initially read `$OUT`, which every control's `run_child` reassigns — so it reported **C15's sandbox** (`ec_ok 19/20` plus C15's own `VIOLATION` line) as the live tree's verdict. A snapshot taken in section A plus `sget_live()` fixes it; sections B and E now read the copy, and E0 asserts the snapshot exists. B was correct only by *ordering* (it runs before any control) — the same latent trap, removed rather than left lucky.
- **`tools/REGISTRY.md`**: the suite's section carries the new count, the E/C14/C15 description and the truth-vs-coverage distinction in its Purpose; **86/86**.

### Verified

- `bash tests/test_registry_coverage.sh` → **86 passed / 0 failed** (~24s, no network), exit 0 — **67 → 86 = +19** (E0–E6 = 7, C14 = 6, C15 = 6), with all 13 pre-existing controls' assertions untouched and still green.
- `tools/regression-run` → **58 suites, 5108 passed, 2 failed, 0 skipped** (pre-commit): **5091 + 19 = 5110 = 5108 + 2**, 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), documented red-before-commit/green-after in `[0.4.107]`.
- The suite caught its own first two defects before any control did: a `sed` range written `/^[*#]s/…` swallows the `s` into the end address and dies with `unknown command: ^`, which reads as an **empty table** — all 20 tools reported undocumented while E1/E2 stayed green — and the `OUT`/`LIVE_OUT` mix-up above. Both are recorded in source comments where the next reader will meet them.

### Notes

**No tool edited**, no service restarted, no certificate touched, no promote executed, **zero DNS writes**, no paid API key, **no secret read or printed**, **spend 0.00** (`*-free` only), `BUDGET.md` untouched at 1.50 / 3.50 for 2026-09. Item **(64)** struck; INBOX line 468 remains open and still blocked on REPORT.md §14 (Nextcloud `{"installed":false}`) — answered in this run's STEP 0 reply (**dev 107 / prod 73**) rather than pretended executed.

### 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
echo "appended"; tail -4 CHANGELOG.md | head -2
appended
**No tool edited**, no service restarted, no certificate touched, no promote executed, **zero DNS writes**, no paid API key, **no secret read or printed**, **spend 0.00** (`*-free` only), `BUDGET.md` untouched at 1.50 / 3.50 for 2026-09. Item **(64)** struck; INBOX line 468 remains open and still blocked on REPORT.md §14 (Nextcloud `{"installed":false}`) — answered in this run's STEP 0 reply (**dev 107 / prod 73**) rather than pretended executed.
> build · mimo-v2.6-flash-free
$ cd /data/repo && ./tools/queue-source-check --format json 2>&1 | head -c 1200; echo; echo "=== repo-lint ==="; ./tools/repo-lint --format json 2>&1 | python3 -c "import json,sys;d=json.load(sys.stdin);print({k:d[k] for k in ('ok','failures','duplicates','citations_missing')}); print('changelog:', {k:d.get('changelog',{}).get(k) for k in ('entries','unique','duplicates')} if isinstance(d.get('changelog'),dict) else d.get('changelog'))"
{"changelog": "/data/repo/CHANGELOG.md", "detail": "one queue: [0.4.114] pointer-only, 111 item line(s) frozen across 71 section(s), 74 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": "0.4.114", "progress_bullets": 74, "queue_headings_fenced": 1, "queue_headings_raw": 72, "queue_sections": 71}, "timestamp": "2026-09-28T15:56:31Z", "tool": "queue-source-check", "violations": []}

=== repo-lint ===
Traceback (most recent call last):
  File "<string>", line 1, in <module>
KeyError: 'duplicates'
> build · mimo-v2.6-flash-free
$ cd /data/repo && ./tools/repo-lint --format json 2>&1 | python3 -c "
import json,sys
d=json.load(sys.stdin)
print('keys:', sorted(d.keys()))
print('ok:', d.get('ok'), 'exit_code:', d.get('exit_code'))
for k in ('failures','errors','checks','summary','changelog_version','citations'):
    if k in d: print(k, '=>', str(d[k])[:600])
"
keys: ['changelog_version', 'error', 'errors', 'exit_code', 'failures', 'files_total', 'go_compile', 'linted', 'ok', 'repo_root', 'requested_sha', 'sha_resolved', 'skipped', 'timestamp', 'tool']
ok: True exit_code: 0
failures => []
errors => []
changelog_version => {'attempted': True, 'path': 'CHANGELOG.md', 'entries': 118, 'unique': 118, 'unparseable': 0, 'duplicates': [], 'citations_seen': 3557, 'citations_in_series': 3433, 'citations_missing': [], 'series': ['0.1', '0.2', '0.3', '0.4'], 'ok': True, 'reason': None, 'summary': '118 changelog version heading(s), 118 unique, 3557 citation(s) checked, 0 missing'}
> build · mimo-v2.6-flash-free
$ cd /data/repo && echo "headings: $(grep -c '^## \[' CHANGELOG.md)"; ./tools/repo-lint --format json | python3 -c "import json,sys;d=json.load(sys.stdin);print('requested_sha',d.get('requested_sha'),'sha_resolved',d.get('sha_resolved'),'files_total',d.get('files_total'),'go_compile',d.get('go_compile',{}).get('ok'))"
headings: 119
requested_sha HEAD sha_resolved b7e78a08a5b88448fcdd76f441da4bc374898fe3 files_total 204 go_compile True
> build · mimo-v2.6-flash-free
$ date -u +%Y-%m-%dT%H:%MZ; cd /data/repo && grep -c "^\- \*\*STEP 0" agent-logs/PROGRESS.md; grep -n "^## 2026-09-28" agent-logs/PROGRESS.md | tail -3
2026-09-28T15:57Z
104
6074:## 2026-09-28T07:2xZ main-loop run — STEP 0 answered FIRST (0 unread both DBs; fresh-probe reply **dev 104 / prod 70**), then **TWO** steps: **(57) EXECUTED** — section **13** compares the go-compile contract's three prose copies, suite **82 → 109** — and **(58)(a) EXECUTED** — the QA entry prepended above the file's own title re-homed `[0.4.29]` → `[0.4.110]` at the bottom, taking **22 of the tree's reds to 0**: regression **57 / 4996 / 0**, *"all suites green"*
6098:## 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**
6113:## 2026-09-28T14:45Z main-loop run — STEP 0 answered FIRST (0 unread both DBs; fresh-probe reply **dev 106 / prod 72**), then ONE step: BOOTSTRAP rules 2-3 made checkable — new `tests/test_registry_coverage.sh`, suite **57 → 58**, `[0.4.113]`
> build · mimo-v2.6-flash-free
$ cd /data/repo && ./tools/inbox-status --format json | python3 -c "
import json,sys
d=json.load(sys.stdin)
def walk(o,p=''):
    if isinstance(o,dict):
        for k,v in o.items():
            if isinstance(v,(dict,list)): walk(v,p+k+'.')
            else: print(p+k,'=',v)
    elif isinstance(o,list):
        print(p+'[len]',len(o))
walk(d)"
tool = inbox-status
timestamp = 2026-09-28T15:57:36Z
sources.inbox = /data/repo/INBOX.md
sources.dev_db = /opt/startup/dev/data/messages.db
sources.prod_db = /opt/startup/prod/data/messages.db
unread.dev = 0
unread.prod = 0
unread.total = 0
unread.rows.[len] 0
last_reply.dev.id = 107
last_reply.dev.ts = 2026-09-28 15:23:12
last_reply.prod.id = 73
last_reply.prod.ts = 2026-09-28 15:23:12
inbox.entries = 74
inbox.handled = 73
inbox.open = 1
inbox.open_replied.[len] 1
inbox.open_unreplied.[len] 0
owed.unread_messages = 0
owed.unreplied_entries = 0
owed.total = 0
ok = True
exit_code = 0
> build · mimo-v2.6-flash-free
$ cd /data/repo && cat >> agent-logs/PROGRESS.md <<'EOF'

## 2026-09-28T16:00Z main-loop run — STEP 0 answered FIRST (0 unread both DBs; fresh-probe reply **dev 107 / prod 73**), then ONE step: **(64) EXECUTED** — every documented exit code 2 is now *probed*, not merely registered: `tests/test_registry_coverage.sh` section **E** + controls **C14**/**C15**, suite **67 → 86**, `[0.4.114]`

- **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 107 / prod 73**, `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 login-verified, 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}` (Nextcloud 34.0.4, never installed), so no admin session exists there and no account can be created yet. The reply restates the one line that closes it (*does `password = own email` extend to Nextcloud?* — a yes/no needs no secret to travel), names what happens the moment it arrives (install, admin-create all seven, never self-registration, a real login probe per account, names-only matrix in `STRUCTURE.md`), names this run's step, and closes with the safety line. **An unanswered investor is a failed run; there was none.**
- **The step, and what it is not**: (64), queued by the `[0.4.113]` entry — *"this suite's `--help` claims exit 2 for unknown arguments and D2 measures it, but the tools side of the same claim (every registered tool's documented exit-2 line vs. its actual behaviour) is unchecked; `run_check` proves `--help` exits 0, not that the table under it is true"*. The fourth BOOTSTRAP field, `Exit codes`, was asserted for **presence** only, so a section could promise `2 - Invalid arguments` while its tool returned `0` for the same argument and every suite in the tree stayed green — the (55)/(57)/(59) shape one child further along, inside the suite written to close that class. Measured first: **20/20** tools answer exit **2** on an unknown flag in ~2s total (validation before any I/O) and **20/20** tables list a code 2, including the two shapes a naive parser misses — `immich-roundtrip`'s **wrapped header** (the `**Exit codes** (precedence …` line continues on the next line) and `smtp-relay-probe`'s **non-sequential** `1 4 3 0 2` (first-match-wins). The invariant held, so this is a **guard, not a repair**: no tool edited, `tools/system-status` byte-unchanged.
- **Two probes in the one code path every control re-execs**: `timeout 5 <tool> --regcov-not-a-real-flag` must exit **2** (`VIOLATION badarg_contract <tool>:rc`), and the section's `**Exit codes**` block must really list a code 2 (`VIOLATION badarg_undocumented <tool>`). The bound is **5s, not the `--help` probe's 20s, and that is the assertion rather than a caveat**: a tool that needs the network or a hang to notice a nonsense argument has failed the rule its own table claims (`dns-verify`'s row spells it out — *"all validation runs before any network or DNS I/O"*). A section with **no** Exit-codes heading stays **C7's** `incomplete_section` and is not re-reported here: one plant, one finding — the new reader owns the table's *content*, the old field check owns its *presence*.
- **The suite caught its own first defect before any control did, and the defect is recorded where the next reader meets it.** The extraction range was written `/^[*#]s/…`, which swallows the `s` into the *end* address: sed died with `unknown command: ^`, the empty result read as *"no table"*, and **all 20 tools reported `badarg_undocumented` while E1/E2 (the behaviour half) stayed green** — 70/15, not 86/0. Fixed to `/^[*#]/s/…` with a source comment naming the symptom, plus the two shapes the end pattern has to survive (wrapped header, non-sequential list), both re-measured by hand against the real sections.
- **The second defect was in my own assertion wiring, not in the tree, and it is the more instructive one.** Section E read `$OUT` — but every control's `run_child` **reassigns `OUT`**, so by the time E ran, `$OUT` was **C15's sandbox** (its `budget-show` table with the `2` bullet deleted): E4 reported `19 of 20` and E6 reported *C15's own `VIOLATION`* as the live tree's verdict. **83/2 with the tree perfectly clean.** Fixed by snapshotting `LIVE_OUT=$OUT` immediately after section A and adding `sget_live()`; E0 now asserts the snapshot exists. Sections **B** were correct only by *ordering* (they run before the first control) — the same latent trap, moved to `sget_live` rather than left lucky. Both bugs are written into the suite's header comment and into `tools/REGISTRY.md`'s *How it stays honest* paragraph, because the second one is a reusable rule: **a live assertion downstream of a control must not read the control's output.**
- **Two new negative controls, one per copy of the claim, each plant-asserted and count-asserted.** **C14** models a tool with no argument validation at all (exit 0 on nonsense, `--help` still correct and its section still present) → `badarg_contract version-check:rc`, **exactly 1** violation. **C15** deletes `budget-show`'s `` - `2` `` bullet with a `re.sub` that asserts `new != body`, then asserts **the tool still exits 2** (only the table changed) → `badarg_undocumented budget-show`, **exactly 1** violation. The behaviour half and the documented half can each be seen going red on the live checker, not on a second checker written for the control.
- **A deliberate edit to two pre-existing controls, disclosed**: C8's and C9's stubs used to fail *every* argv, so with the new probe each plant would be caught **twice** (`help_failed` + `badarg_contract`) and `exactly one violation` would be measuring the plant's sloppiness instead of the checker's precision. Both stubs now break **only `--help`** (exit 3 / silent exit 0 there, correct exit 2 elsewhere) with a source comment saying why; their own plant assertions (C8a `--help` exits 3, C9a/C9b `--help` exits 0 and is empty) are unchanged, as are all 13 controls' assertion texts.
- **Measured — the arithmetic closes**: `bash tests/test_registry_coverage.sh` → **86 passed / 0 failed**, exit 0, ~24s (was **67**; **+19 = E0–E6 (7) + C14 (6) + C15 (6)**), standalone. `tools/regression-run --log-dir /tmp/opencode/regression-0.4.114` → **58 suites, 5108 passed, 2 failed, 0 skipped** (pre-commit): **5091 + 19 = 5110 = 5108 + 2**, the arithmetic that says the step added only itself, and the 2 reds are `test_gladex_monitor.sh`'s dirty-tree pair **A3**/**A15**, the pair `[0.4.107]` documents as red-before-commit/green-after — the only failing suite in the run, confirmed by name rather than by count. Gates read **after** the edits, none carried from before them: `tools/queue-source-check --format json` → **exit 0, `violations []`**, `[0.4.114]` pointer-only, **111** frozen item lines across **71** sections, `newest_queue_items 0`, `progress_bullets 74`; `tools/repo-lint --format json` → **exit 0, `ok true`, `failures []`, `errors []`, `go_compile.ok true`**, `changelog_version.entries 118 / unique 118 / duplicates [] / citations_missing []` — read at **HEAD `b7e78a0`**, so the 119th heading this run appended is counted only after the commit, as every previous entry notes.
- **`tools/REGISTRY.md`'s own suite section carries the truth-vs-coverage distinction**: Purpose now says *since `[0.4.114]` it also asserts the **truth** reading of that fourth field*, the *How it stays honest* paragraph gains the 15-control count and the `LIVE_OUT` rule, the Tests block lists E and both new controls, and **Status** reads **86/86**. No other registry `Status` line was touched — the stale-`52/4267` figure is still item **(63)**, deliberately not opened here.
- **Deliberately not done**: **only code 2 is probed** — the other documented codes (`3` unreachable, `4` expiring, `5` rejected, `7` refused …) need a live service, a relay or DNS, and are queued as **(65)** rather than half-probed offline; the bullet's *text* is likewise read for the number, not compared against `--help`'s wording (**(66)**, the (55) pattern, compare-or-declare (54)/(56)). **No tool edited**, no `tools/` script behaviour changed, **no version-train bump** (`GLADEX_APP_VERSION` stays `0.4.28`), **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 listed exactly this run's files (`tests/test_registry_coverage.sh`, `tools/REGISTRY.md`, then `CHANGELOG.md` and this file), 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), (60), (61), (62) and (63) from the 14:45Z entry unchanged — item **(64) struck — EXECUTED by `[0.4.114]`** (20 tools, 20 tables, zero readers; two probes in `run_check`, section E, controls C14/C15, the `sed` range and `OUT`/`LIVE_OUT` bugs both self-caught and both written into the source). New from this run: **(65)** E proves the **code 2** line only — every other documented code in the 20 tables (`tls-check`'s `4`/`5`, `promote-dev-to-prod`'s `5`–`8`, `healthcheck`'s `3` …) still has **no reader**, and most cannot be probed without a live service/relay/DNS, so the compare-or-declare choice applies to them too: derive the offline-reachable subset (a bad *value* of a known flag — `--format xml`, `--timeout 0`, a malformed domain — which several tables promise to reject before I/O) and probe that, or declare in `REGISTRY.md` that only code 2 is machine-checked. **(66)** the documented half checks that a `` - `2` `` bullet **exists**, not what it says: a table reading `` - `2` - internal retry count `` would pass E4 green, exactly the (55) failure mode on a new field — either match the bullet's wording against the tool's own error line (or `--help`'s) or state the granularity in the section's comment, compare-or-declare as in (54)/(56).
- **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 107 / prod 73).
EOF
echo appended; ./tools/queue-source-check --format json | python3 -c "import json,sys;d=json.load(sys.stdin);print('ok',d['ok'],'violations',d['violations'],'rules',{k:d['rules'][k] for k in ('newest_version','frozen_items_actual','progress_bullets','newest_queue_items','queue_sections')})"
appended
ok True violations [] rules {'newest_version': '0.4.114', 'frozen_items_actual': 111, 'progress_bullets': 75, 'newest_queue_items': 0, 'queue_sections': 71}
> build · mimo-v2.6-flash-free
$ cd /data/repo && for t in tests/test_queue_source.sh tests/test_repo_lint.sh; do printf '%-34s ' "$t"; out=$(bash "$t" 2>&1 | tail -1); echo "$out"; done; for t in tests/test_changelog_api.php tests/test_changelog_mobile.php; do printf '%-34s ' "$t"; out=$(php "$t" 2>&1 | tail -1); echo "$out"; done
tests/test_queue_source.sh         ===============================
tests/test_repo_lint.sh            ===============================
tests/test_changelog_api.php       === Results: 86 passed, 0 failed ===
tests/test_changelog_mobile.php    === Results: 125 passed, 0 failed, 0 skipped ===
> build · mimo-v2.6-flash-free
$ cd /data/repo && for t in tests/test_queue_source.sh tests/test_repo_lint.sh; do printf '%-34s ' "$t"; bash "$t" 2>&1 | grep -E "Results|passed," | tail -2 | tr '\n' ' '; echo; done
tests/test_queue_source.sh         queue-source-check suite: 181 passed, 0 failed 
tests/test_repo_lint.sh            repo-lint suite: 420 passed, 0 failed
> build · mimo-v2.6-flash-free
$ cd /data/repo && git status --porcelain && git add tests/test_registry_coverage.sh tools/REGISTRY.md CHANGELOG.md agent-logs/PROGRESS.md && git -c user.name="Atlas" -c user.email="atlas@gladex.de" commit -q -m "[0.4.114] queue item (64): every documented exit code 2 is now probed, not just registered — section E + controls C14/C15, suite 67 -> 86" && git log --oneline -2 && git status --porcelain
M CHANGELOG.md
 M agent-logs/PROGRESS.md
 M tests/test_registry_coverage.sh
 M tools/REGISTRY.md
0b1f36b [0.4.114] queue item (64): every documented exit code 2 is now probed, not just registered — section E + controls C14/C15, suite 67 -> 86
b7e78a0 [0.4.113] post-commit confirmation: 58 suites / 5091 / 0 failed, all suites green
> build · mimo-v2.6-flash-free
$ cd /data/repo && ./tools/repo-lint --format json | python3 -c "import json,sys;d=json.load(sys.stdin);c=d['changelog_version'];print('repo-lint ok',d['ok'],'exit',d['exit_code'],'failures',d['failures'],'entries',c['entries'],'unique',c['unique'],'duplicates',c['duplicates'],'citations_missing',c['citations_missing'],'go_compile',d['go_compile']['ok'])"; ./tools/inbox-status --format json | python3 -c "import json,sys;d=json.load(sys.stdin);print('inbox ok',d['ok'],'owed.total',d['owed']['total'],'unread',d['unread']['total'],'open',d['inbox']['open'],'open_unreplied',len(d['inbox']['open_unreplied']),'last dev',d['last_reply']['dev']['id'],'prod',d['last_reply']['prod']['id'])"; ./tools/system-status --format json | python3 -c "import json,sys;d=json.load(sys.stdin);print('system-status rc',d.get('exit_code'),'overall',d.get('overall'),'errors',len(d.get('errors',[])) if isinstance(d.get('errors'),list) else d.get('errors')); [print('  WARN:', w.get('check') or w) for w in (d.get('warnings') or [])]"
repo-lint ok True exit 0 failures [] entries 119 unique 119 duplicates [] citations_missing [] go_compile True
inbox ok True owed.total 0 unread 0 open 1 open_unreplied 0 last dev 107 prod 73
system-status rc None overall ok errors 0
> build · mimo-v2.6-flash-free

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

[Mon Sep 28 18:02:59 2026] 127.0.0.1:57162 Accepted
[Mon Sep 28 18:02:59 2026] 127.0.0.1:57162 Closing
[Mon Sep 28 18:03:35 2026] 127.0.0.1:35932 Accepted
[Mon Sep 28 18:03:35 2026] 127.0.0.1:35932 Closing
[Mon Sep 28 18:03:35 2026] 127.0.0.1:35942 Accepted
[Mon Sep 28 18:03:35 2026] 127.0.0.1:35942 Closing
[Mon Sep 28 18:03:50 2026] 127.0.0.1:47414 Accepted
[Mon Sep 28 18:03:50 2026] 127.0.0.1:47414 Closing
[Mon Sep 28 18:04:21 2026] 127.0.0.1:51670 Accepted
[Mon Sep 28 18:04:21 2026] 127.0.0.1:51670 Closing
[Mon Sep 28 18:04:21 2026] 127.0.0.1:51684 Accepted
[Mon Sep 28 18:04:21 2026] 127.0.0.1:51684 Closing
[Mon Sep 28 18:04:21 2026] 127.0.0.1:51696 Accepted
[Mon Sep 28 18:04:21 2026] 127.0.0.1:51696 Closing
[Mon Sep 28 18:04:21 2026] 127.0.0.1:51700 Accepted
[Mon Sep 28 18:04:21 2026] 127.0.0.1:51700 Closing
[Mon Sep 28 18:04:21 2026] 127.0.0.1:51708 Accepted
[Mon Sep 28 18:04:21 2026] 127.0.0.1:51708 Closing
[Mon Sep 28 18:04:37 2026] 127.0.0.1:55650 Accepted
[Mon Sep 28 18:04:37 2026] 127.0.0.1:55650 Closing
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55662 Accepted
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55662 Closing
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55674 Accepted
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55674 Closing
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55690 Accepted
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55690 Closing
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55698 Accepted
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55698 Closing
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55714 Accepted
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55714 Closing
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55724 Accepted
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55724 Closing
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55726 Accepted
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55726 Closing
[Mon Sep 28 18:04:38 2026] 127.0.0.1:55740 Accepted
[Mon Sep 28 18:04:39 2026] 127.0.0.1:55740 Closing
[Mon Sep 28 18:04:39 2026] 127.0.0.1:55752 Accepted
[Mon Sep 28 18:04:39 2026] 127.0.0.1:55752 Closing
[Mon Sep 28 18:04:39 2026] 127.0.0.1:55768 Accepted
[Mon Sep 28 18:04:39 2026] 127.0.0.1:55768 Closing
[Mon Sep 28 18:04:39 2026] 127.0.0.1:55782 Accepted
[Mon Sep 28 18:04:39 2026] 127.0.0.1:55782 Closing
[Mon Sep 28 18:04:39 2026] 127.0.0.1:55784 Accepted
[Mon Sep 28 18:04:39 2026] 127.0.0.1:55784 Closing
[Mon Sep 28 18:04:39 2026] 127.0.0.1:55792 Accepted
[Mon Sep 28 18:04:39 2026] 127.0.0.1:55792 Closing
[Mon Sep 28 18:04:39 2026] 127.0.0.1:55806 Accepted
[Mon Sep 28 18:04:39 2026] 127.0.0.1:55806 Closing
[Mon Sep 28 18:04:49 2026] 127.0.0.1:53666 Accepted
[Mon Sep 28 18:04:49 2026] 127.0.0.1:53666 Closing
[Mon Sep 28 18:05:50 2026] 127.0.0.1:38908 Accepted
[Mon Sep 28 18:05:50 2026] 127.0.0.1:38908 Closing
[Mon Sep 28 18:06:50 2026] 127.0.0.1:39942 Accepted
[Mon Sep 28 18:06:50 2026] 127.0.0.1:39942 Closing
[Mon Sep 28 18:07:50 2026] 127.0.0.1:47342 Accepted
[Mon Sep 28 18:07:50 2026] 127.0.0.1:47342 Closing
[Mon Sep 28 18:10:05 2026] 127.0.0.1:54900 Accepted
[Mon Sep 28 18:10:05 2026] 127.0.0.1:54900 Closing
[Mon Sep 28 18:10:48 2026] 127.0.0.1:37898 Accepted