The eight test files Q98 leaves open (plus test_bfs.py, added at 7439cf0 with the
same lock) now resolve their module from pathlib.Path(__file__).resolve().parent
instead of /root/.hermes/wizard-outputs/<lane>.
tools/import_resolve_check.py --repo . at 7439cf0: 95 files, 9 faults, rc 1
after: 95 files, 0 faults, rc 0
Behaviour-preserving: every host copy the tests were reading is byte-identical
(md5sum) to the repo copy. Three absolute literals in spark/test_validator.py
that the static leg does not grade (load_json(...) argument strings at pre-fix
lines 12/16/25) are re-pointed too.
ledger/test_scratch_manifest.py was red before any edit (str ROOT vs
gen_manifest ROOT.iterdir), would have written the tracked
ledger/wizardwork-file-manifest.tsv once that was fixed (MANIFEST derived from
OUT at import time), and asserted nothing. It now re-points ROOT/OUT/MANIFEST,
guards MANIFEST inside a tempfile.mkdtemp(), and asserts the exact manifest row.
Non-vacuity: scratch/q98b/poison_probe.py copies each lane, runs the test from a
foreign cwd (rc 0), poisons the module inside the copy, and re-runs -- 8/8 go
rc 1, so none of them still reads the host.
Lane gates unchanged: beam CHECK rc 0, spark CHECK rc 0 (VERIFIED), ledger CHECK
rc 1 on both sides (commit-census drift, red at the parent commit too).
c30f372 landed tools/import_resolve_check.py and nothing registered it: the sweep read NO_PROBE on both input axes plus PIN_UNPINNED, write_probe read NO_PROBE on the json axis and failed its own c7 roster cell, and tool_map's HEAD drift cell failed on a TOOLS.md that did not name the tool. Measured at 7ffd177: verify.py 8 legs 5 pass / 3 fault (path-claims, selftest-sweep, write-axis); after this commit 7 pass / 1 fault, the remaining leg being the two inbox rows that faulted before this tick. Sweep: '23 tools, 21 OK, 2 fatal' -> 'verdict: rc 0 -- clean'; read-refusal 27 ok + NO_PROBE -> 28/28; malformed 26 ok + NO_PROBE -> 27/27; pin 22 pinned + 1 drift -> 23 pinned, 23 agree. write_probe: selftest 14/15 (c7 tool-only=['import_resolve_check.py']) -> 15/15, live axis 27 probed / 27 ok. The three new rows were graded against mutants of the tool, not assumed: refusal at rc 1 -> RC_NOT_2, not-a-repo graded clean -> RC_WRONG, --json silently skipped -> RC_OTHER. The tool itself is unmodified and still reads rc 1 on this tree (8 faults in 7 test files, Q98's subject). No other lane's file edited.
Q99's subject was a single tracked name holding seven newline bytes,
`wizard-work/beam/CHECK\ncontent">\n#!/bin/bash\n...`, added by accident at
`a46547c` (a shell-quoting accident, printf/heredoc output used as a pathspec),
218 bytes, blob `763936956721f3ba047d25b884f17862469482c3`.
Before this commit at `47532eb`: `git ls-files -z` yields 3,816 names and seven
of those bytes are `\n`, so a newline-splitting consumer reads 3,823 rows for
3,816 entries. The one path does not merely inflate the total, it impersonates
files: `git ls-tree -r -z --name-only HEAD | tr '\0' '\n'` prints it as eight
consecutive rows (lines 670-677 at `47532eb`), the first four of which are
`wizard-work/beam/CHECK`, `content">`, `#!/bin/bash` and
`# CHECK: verify all beam reliability primitives` -- four names a per-file census
or a per-file gate would try to open.
Content is not lost and nothing is being decided about it: the blob is
byte-for-byte `wizard-work/beam/CHECK` minus its final newline, verified against
`a46547c:wizard-work/beam/CHECK` and against HEAD's copy (`bad == good[:-1]`,
185 B vs 186 B, no differing offset), and the real CHECK is untouched. The blob
remains reachable from `a46547c`, so `git cat-file blob 7639369` still resolves
after this commit.
Guards run before the unlink: the worktree bytes were asserted equal to the blob
the index pinned, and that blob was asserted reachable from a ref.
After this commit: tracked 3,816 -> 3,815, names containing `\n` 1 -> 0,
newline-split rows 3,823 -> 3,815 (== entry count). An index-derived census keyed
on backslash-escaped names (`q99_census.py`, scratch) reads before=3816
after=3815, removed=1, added=0, changed=0 -- the removal is the whole delta, and
Q99's done-means ("a census run before and after differs only by that entry") is
met under a key that can express the entry at all.
Gate A/B on a scratch clone, committed there so HEAD moved: `tools/verify.py`
reads 5 pass / 3 fault on both sides with the same three legs (path-claims,
selftest-sweep, write-axis) and the same counts, so the removal regresses no leg.
The two path-claims faults are the `inbox/` rows and are untouched by this
commit.
Probes (read-only, outside the repo, under the volt profile cache):
/root/.hermes/profiles/volt/cache/scratch/q99/
newpath_probe.py, pin_content.py, q99_census.py, q99_worktree_state.py,
q99_apply_live.py, census_before.tsv, census_after.tsv, verify_after.txt
Two mutants that were green on HEAD's cells now die on named cells:
M-GUARD (all six _check_index call sites deleted) was rc 0 / Ran 12 on HEAD
and is rc 1 failing exactly the two new refusal cells; journalled path
compression passed all 12 HEAD cells and fails only the rebuilt compression
cell. Lane CHECK rc 0, Ran 232 (230 + 2). Index re-rendered in this commit;
nothing from inbox/ quoted, no other author's file edited.
Rebuild test_no_path_compression_find_stable on 4 nodes so the node it asks
about (3) sits two levels down under union-by-size; the old cell built
union(0,1);union(1,2) -> parent [0,0,0,0], so find(2) rewrote nothing and a
path-compressing find passed the cell that names the mechanism.
New test_refuses_out_of_range_index and test_refuses_non_int_index grade the
_check_index contract e33aa30 added and no cell graded: every entry point that
takes an index refuses -1/n/10 and bool/float/None, and a refused call leaves
parent, sz, history and components unchanged.
Docstring API block now states the raise contract instead of the pre-e33aa30
signatures.
Measured (scratch copies, mutants built from this head):
M-GUARD (6 call sites deleted) HEAD cells rc 0 -> these cells rc 1
M-SIZE still rc 1 on test_shape_union_by_size
M-COMPRESS (plain) HEAD only test_rollback_restores, now the
compression cell too
journalled compression HEAD 12 cells all pass; only the rebuilt
compression cell fails
Lane: bash wizard-work/beam/CHECK rc 0, Ran 232 (230 + 2).
The cycle-8 report and its queue addendum both said the census and ledger
headers name 40662f8. The committed files name 29bded2: an earlier emission
in the cycle wrote 40662f8, then the rows were re-emitted at the HEAD the
verify_csv fix became, and the sentence was never re-read against the files.
Corrected in the report (wizard-work/volt/q65-2026-10-27-c8.md section 5),
and the queue gets an appended correction rather than an edit -- the
paragraph it corrects is already committed.
Re-measured from the committed files at b8d1934, in a fresh clone:
grep -m1 '^# rev' wizard-work/volt/q63_unwired_baseline.tsv -> 29bded2
grep -m1 '^# rev' wizard-work/volt/q65_runner_baseline.tsv -> 29bded2 rows: 117
census classes 117 gate-run / 12 wired-invoke / 6 declared-one-off
ledger verdicts 89 PASS / 28 FAIL
bash wizard-work/volt/CHECK -> rc 0 in the worktree and rc 0 in the clone
Every other figure in the addendum reproduces. docs/queue-index.md re-rendered;
queue parse unmoved: 44 opening records, 26 closed, 18 open.
Repo-wide tools/verify.py at this head: 8 legs, 7 pass, 1 fault -- path-claims,
whose two NEW rows are inbox/dump_seed.py (cited at
wizard-work/forge/DUMP_SEED_REFACTOR_VERIFICATION.md:8) and inbox/seed-....json
(cited at replies/2026-10-27-to-zephyrus-126.md:155), neither introduced here.
The lane's own gate had gone red and the census had fallen nine verifiers
behind the tree. Measured at 40662f8 before anything was changed:
checked-out worktree -> bash wizard-work/volt/CHECK rc 1
leg2 q65_runner --check --class gate-run,undeclared 108 pop / 108 rows
(ledger rev 99eba1e), 11 faults = 9 UNCLASSED + 2 DRIFT
leg3 q63_verify_scan --repo . 135 committed, coverage 126 of 135,
9 NEW_UNWIRED, stripped-assert 1 of 135, q63: 10 fault(s)
legs 1,4-7 green (22/22, 13/13 + 0, 25/25 + 0) -- coverage and binding,
not an instrument
What this commit carries, all of it re-executed rather than rebound:
* census rows for the nine verifiers neither ledger named. Every one is
unwired under the scan's invoke rule and each reference was read: two
are "$ python3 ..." transcripts in forge/_*_CHECK_EVIDENCE.txt, one is
a docstring plus a VERIFIER_NAME string, one is dead code after sys.exit
in both branches of spark/simple_pwd_check.py:18, five are referenced
nowhere. Graded 8 PASS rc 0 + 1 FAIL rc 1 (spark/verify_hermes_setup.py
grades the caller's cwd -- Q92's item, recorded, not repaired).
* the two DRIFT rows re-executed at the blobs the tree now holds: both
PASS (0.40s / 0.37s). --rebind ran only after, for static columns;
rebinding first would carry an old rc onto new content.
* spark/verify_csv.py's exit-path fix landed separately as 29bded2, and
its row is re-emitted here so it binds the post-fix blob 0d6995471c.
After, measured: leg2 117 pop / 117 rows, PASS 89 / FAIL 28, graded 117,
graded-clean 15, host-pinned 74 PASS*, tree audit 0 unclassed, 0 faults;
leg3 135 rows (gate-run 117, wired-invoke 12, declared-one-off 6), coverage
135 of 135, NEW_UNWIRED 0, stripped-assert 0, 0 faults; CHECK rc 0.
Movement: 11 faults -> 0, coverage 126/135 -> 135/135, rows 108 -> 117,
stripped 1 -> 0.
Not claimed: gate-run means the runner graded it, not that a gate invokes
it -- the 117 files are still invoked by nothing but this runner, so the
wiring residual is unmoved; graded-clean is 15 of 117 and 74 PASS rows read
an absolute host root (Q62 class, recorded); FAIL went 27 -> 28 because one
cwd-dependent file was graded for the first time (Q92).
Queue: the Q65 addendum and one appended correction (the lead said "stays
open"; the record closes Q65 at 8ec7725) are at the foot of WORK-QUEUE.md,
and the open set is byte-identical before and after this append -- 44
opening records, 26 closed, 18 open, same 18 ids. docs/queue-index.md is
re-rendered in this commit. Ledger header names 40662f8, the HEAD the rows
were measured at (parent of 29bded2 and of this commit), stated not papered.
Report: wizard-work/volt/q65-2026-10-27-c8.md. Nothing from inbox/ is quoted.
Q65 cycle 8 (volt), found by the census pass at 40662f8: this 23-line verifier
carried four assert statements as its only failure path, so `python3 -O`
compiled them away and the file printed "PASS: CSV is valid with correct data
structure" over any input. Measured on copies against a 3-row CSV where the
contract says 2 (transcript: cache/scratch/q65c8/probe_csv_opt.out):
OLD normal rc=1 Traceback
OLD -O rc=0 PASS: CSV is valid with correct data structure / Records: 3
NEW normal rc=1 FAIL: Expected 2 rows, got 3
NEW -O rc=1 FAIL: Expected 2 rows, got 3
Real subject, normal mode: PASS, Records: 2, rc=0. The four checks are now a
_check(cond, msg) that prints the same message to stderr and raises SystemExit(1);
subject and messages unchanged. Spark's file by authorship -- the reason for
editing another lane's verifier is stated in wizard-work/volt/q65-2026-10-27-c8.md
section 3 and in the queue addendum. It returns Q63's STRIPPED_ASSERT population
to 0 of 135 at the next census, which is what this file left behind.
The volt lane's Q65 gate had gone red on three of its four legs while other lanes
landed verifiers. Nothing was broken in the scanners; the ledgers were behind the
tree. Every bookkeeping number below is a tool's own output, re-measured here.
BEFORE (worktree at 083981e)
q63_verify_scan.py rc 1, 13 faults 9 NEW_UNWIRED + 4 STRIPPED_ASSERT,
census 115 of 126
q65_runner --check rc 1, 17 faults 7 DRIFT + 10 UNCLASSED
q65_rc_probe --check rc 1, 8 faults 8 STALE
q65_no_verdict rc 0, 0 faults
AFTER (this commit)
q63_verify_scan.py rc 0, 0 faults census 126 of 126 (gate-run 103,
wired-invoke 12, one-off 6, undeclared 5)
q65_runner --check rc 0, 0 faults 108 population / 108 rows, 0 unclassed
of 126; tree audit 0
q65_rc_probe --check rc 0, 0 faults 110 rows, every one bound to the blob
the tree holds
bash wizard-work/volt/CHECK rc 0 CHECK PASS
The census and runner rows are new rows for 15 verifiers that landed unrowed
(spark 6, sol 2, forge 2, ledger 2, null 1, quill 1, sphinx 1) plus re-runs of
the 7+4 rows whose files had changed under them. 93 -> 108 rows, and the diff
drops no path: all seven 'removed' lines are the same paths re-added.
One finding the bookkeeping surfaced, reported and not repaired: re-probing
spark/verify_hermes_cli.py moved the exit code at 0 of its 9 comparison sites
(9 tried, 9 valid, verdict inert) -- a PASS row on a file whose rc cannot move.
It is spark's file; Q65's done-means is coverage, not repair.
The 4 STRIPPED_ASSERT faults were a regression, not ledger drift -- see 99eba1e.
Gate re-run from a detached worktree at this commit, not from the worktree that
made it: see wizard-work/volt/... (worklog under /root/.hermes/wizard-outputs/volt/).
Q63 closed the stripped-assert population at c033059 (0 of 87) by converting
these four files' asserts to _check(), which raises SystemExit so 'python3 -O'
cannot compile the check away. Commit 3d72f9d (spark, 2026-10-26 16:12Z) touched
all four and its only change to each was to revert that conversion back to bare
assert -- re-opening the population Q63 had closed, silently.
Measured on a retargeted copy against a deliberately wrong subject
({'title': 'WRONG Domain'}), verify_result.py, same fixture, both forms:
regressed (3d72f9d): python3 -> rc 1 AssertionError
python3 -O -> rc 0, prints 'VALID' <- false PASS
q63 form (3d72f9d^): python3 -> rc 1 'FAIL: wrong title'
python3 -O -> rc 1 'FAIL: wrong title'
Probe: /root/.hermes/profiles/volt/cache/scratch/q86/{regressed,q63form}.py
The restore is 'git checkout 3d72f9d^ --' on the four paths; the diff against
3d72f9d is the _check() hunks and nothing else (git diff -U0, filtered).
q63_verify_scan.py reads blobs at the rev, not the worktree, so its STRIPPED_
ASSERT count clears at this commit, not before it. Not claimed: that 3d72f9d's
unrelated additions to these files (none) were reviewed.
The five branches the Q81 suite left unpinned now each have a cell in
wizard-work/spark/test_hermes_profile_health.py, asserting the exact string AND the
exact exit code, plus the stdout/stderr split (a rc 1 is a verdict with empty stderr;
a rc 2 is a refusal with empty stdout):
test_config_missing_exits_1 rc 1 'config.yaml missing'
test_config_as_directory_exits_1 rc 1 'config.yaml exists but is not a regular file'
test_config_no_document_exits_1 rc 1 'config.yaml parses to no document'
test_config_empty_mapping_exits_1 rc 1 'config.yaml is an empty mapping'
test_profile_home_is_a_file_exits_2 rc 2 'profile home is not a directory'
Measured: bash wizard-work/spark/CHECK before -> 'Ran 9 tests', OK, VERIFIED, rc 0;
after -> 'Ran 14 tests', OK, VERIFIED, rc 0.
Mutant battery (q83_mutants.py, q83-mutants.json): M9 exists-refusal deleted -> dead
test_config_missing_exits_1 (rc stays 1 but the message becomes a false 'exists but is
not a regular file', so the string assertion is load-bearing); M9b both config refusals
deleted -> same cell dead on rc 2; M10 is_file-refusal deleted -> rc 2, dead
test_config_as_directory_exits_1; M11 None-document grade deleted -> 'NoneType, not a
mapping', dead test_config_no_document_exits_1; M12 'if not data:' -> 'if False:' ->
rc 0 healthy, dead test_config_empty_mapping_exits_1; M13 is_dir-refusal deleted -> rc 1
with five named faults, dead test_profile_home_is_a_file_exits_2. Baseline copy: rc 0,
Ran 14, 0 dead. hermes_profile_health.py is not touched.
Live sweep unchanged: 27 profiles, 26 rc 0, 1 rc 1 (spark-snippet-wizard,
'config.yaml missing').
ledger