Round-8 review (Claude, register key `lift-guard-regex-both-directions`).
Every class reproduced here before fixing, and re-tested after.
THREE FALSE NEGATIVES — the payload reads a name and the guard stayed silent,
which is the direction that costs something, because silence is what the tool
exists to prevent:
echo $((X + 1)) arithmetic expansion reads X without a `$` before the
(( Y > 0 )) name, and the read pattern cannot match it: the
character after `$` is `(`. Both contexts are now
tokenised. `if [ $((inm + ins)) -eq 0 ]` is already live
at check.sh:464 — not lifted today, so latent, not absent.
n=Q; ${!n} indirect expansion defeats text analysis outright. The
guard now REFUSES the lift rather than passing it. Its
contract is "does not miss a dependency"; where it cannot
honour that it must say so, not shrug.
SIX FALSE POSITIVES — the driver defines the name and the guard cried wolf.
This direction matters too: a guard that raises false alarms gets edited away,
and then it guards nothing.
case x in a) FOO=1 ;; `)` added to the assignment delimiters
if …; else FOO=1; fi `else` added
! FOO=1 `!` added
mapfile -t FOO binds a name with no `=` at all
readarray -t FOO likewise
printf -v FOO "x" likewise
The banner also over-claimed. It read as a completeness statement about
LIFTING; it is a completeness statement about VARIABLES. A lifted payload also
inherits functions, shell options, traps and a working directory, and this tool
models none of them — loud failures under `set -e`, but the header now says so
rather than implying otherwise.
Verified: all nine classes behave correctly, a genuine missing variable is
still caught by name, and the seven lifting self-tests pass in dalek plus the
four fast ones in each ported fork.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Closes four round-7/8 findings. Certified by the round-12 sweep: five
repositories, both buttons and every self-test, 48/48 GREEN.
── `scalar-statements-unbound` (gpt, round 7, CRITICAL) ────────────────────
The main button bound its 31 certificates' elaborated statements and reachable
specification bodies. This button bound NONE of its thirteen, while
TRUSTED-BASE item 8 said the audit covers "every certificate" — false across
the 44-certificate surface. The finding was raised in round 7, lost from the
round-8 work list by an F-number collision between two reviewers, and re-raised
in round 8.
Proofs/ScalarAudit.lean is generated from each fork's OWN Audit.lean, so the
canonicalisation is provably the same code: pp.all rendering, whitespace
normalisation, transitive specification closure. check-scalar.sh Phase 3c pins
the block's digest, requires the committed copy to match byte-for-byte so a
mismatch can be DIFFED, and cross-checks the auditor's certificate set against
the button's CERTS array.
dalek ecf3a3f8 · anza 0d942e47 · risc0 4b550a61 · betrusted 4b550a61
risc0 and betrusted share a digest and that is correct, not a collision: their
ScalarSubSpec.lean differs only in doc prose and in `black_box` entries inside
`simp only [...]` lists AFTER `:= by`. Proof scripts. They bind the same
statements over the same specifications, which is the documented scope.
selftest-scalar-statements.sh ships the two attacks the reviewer asked for:
ok gutted statement caught (cone unchanged)
ok rewritten specification body caught (name and cone unchanged)
The second rewrites a reachable reference body to `id (…)` — DEFINITIONALLY
EQUAL, so the corpus compiles and every proof typechecks and the cone is
byte-identical. Every earlier phase is blind to it.
── `drv-surface-no-cones` + `accounting-certifies-enumeration` (claude) ────
The round-7 accounting identity proved every kernel constant was ENUMERATED.
The reviewer showed enumeration is not audit: their planted claim WAS
enumerated, as DRV|LTLAccAudit.bait.smuggled|theorem with a real cone, and
nothing examined it — rows had no cone, no allowlist covered them, the
statement digest does not reach instruments, and Phase 2b gates DECLARED
AXIOMS, a different question. "Progress of one step, not two."
DRV rows now carry their axiom cone and are pinned in driver-allowlist.txt by
inventory_gate.sh with a DRV tag — the same implementation that pins the
corpus, in both directions, because a second copy of a coverage gate is a
second thing to drift. The axiom policy is per-surface and enforced per
surface: the corpus admits exactly the sanctioned boundary, the instruments
admit none, and an instrument axiom fails EVEN WHEN ALLOWLISTED.
Verified with the reviewer's own payload, both placements:
before the walk -> UNCLASSIFIED: DRV|…|bait.smuggled|theorem|Classical.choice,Quot.sound,propext
after the walk -> ACCOUNTING FAILED names it (kernel-side)
── `drv-naming-heuristic` (claude, round 7) ────────────────────────────────
Retired as load-bearing rather than patched. The rule admits a theorem whose
name extends a constant declared alongside it, and "breaks in one line" —
declare `def bait`, then `theorem bait.smuggled` walks through. It stays as a
fast readable first check; membership in a committed allowlist is what now
carries the weight, and a new row fails closed whatever it is called.
── what round 11 caught, which was mine ───────────────────────────────────
DRV rows first shipped WITHOUT their originating driver. dalek and anza run
two drivers, each declaring its own `corpus`; keyed on name alone those two
distinct declarations produced one byte-identical row, `sort -u` collapsed
them, and the trailers summed to 37 against 36. The estate had already learned
this on the corpus walk — INV rows carry their module because two modules both
declare CurveFieldProofs.zero_spec — and I rebuilt the record without it.
Rows now carry their driver, and the gate FAILS CLOSED ON DUPLICATE RECORDS
naming the collision: two declarations sharing one entry means one is covered
by the other's, which is exactly how a real declaration hides. The trailer
now checks what the drivers EMITTED, not what survives de-duplication —
conflating "the run was truncated" with "two rows were identical" is what let
a record-format defect present itself as an arithmetic complaint.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Round-8 review (GPT-5.6, register key `section-prefix-bug`, CRITICAL).
Reproduced here exactly before fixing.
model-correspondence.py treated `namespace`, `section` and `end` as one event
class and pushed a named section onto the fully-qualified-name prefix. Lean
does not: `section Foo` opens a scope for `variable`/`open` and gives `end Foo`
a label; it does not turn `bar` into `Foo.bar`. Given a template reading
section Foo
axiom bar : Nat
end Foo
the scanner reported `Foo.bar`, `--names` handed Phase 2d only `Foo.bar`, Lean
resolved an unrelated `Foo.bar` definition elsewhere in the corpus, and the
verdict came back PROVEN. The axiom the extraction ACTUALLY depends on was
never queried. This survived both the fail-closed rewrite and the new
Lean-semantic phase, in a scanner rewritten that same week specifically to
stop dropping things.
AND THE REASON IT STAYED SILENT, which is the half worth keeping. The real
external did not vanish — it landed in the table as EXTRA, the one verdict
that could not fail. A silent bucket beside a fail-closed parser is a slower
way of dropping things. An extra AXIOM is now EXTRA-AXIOM and stops the
button: the model exists to answer the template, so an assumption nothing
asks for is either a parse we got wrong or an assumption nobody governs.
Extra definitions stay tolerated; helpers in a model file are ordinary.
That gate fired on the real corpora on its first run. Each fork's
hand-maintained gen/CurveField/FunsExternal.lean carried AVX2/AVX512 backend
axioms present in no template, no proof, no cone and no allowlist — dead
assumptions in a pinned trusted-base file, reported as EXTRA and therefore
invisible. extract.sh:16 confirms these files are never overwritten by
extraction, so they were hand-written and are removed here:
dalek 2, anza 3, risc0 4, betrusted 4
Nothing referenced them, so no certificate's cone changes; the trusted base
simply gets smaller. Table rows 64->62, 51->48, 57->53, 56->52, and Phase 2d
independently resolved 62/48/53/52 externals against the regenerated tables.
GEN-MODEL.sha256 and HARNESS.sha256 both move: the model bytes changed, and
the harness pins the table and the gen manifest themselves.
Certified: round-10 sweep, 2h53m, ten instruments in each of four forks,
40/40 GREEN, 0 failing, 0 resource-limited. A full run was required — the
--audit-only staleness gate correctly refused after a source change.
Registered in formal-verification-control/review-findings.tsv as
`section-prefix-bug` and `dead-model-axioms`.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Round-9 swept all ten instruments per fork instead of only the two buttons,
and found selftest-axgate.sh red in all four. Root cause, identical to the
KERNLOG defect fixed the day before: a self-test lifts one phase out of
check.sh and supplies its variables from a HAND-WRITTEN preamble, and the
phase later grew a dependency nobody told the preamble about.
Phase 2b stopped globbing Proofs/*.olean and started reading $PROOFS by
membership. Bash does not error on an unset array under `set -u` — it
expands to nothing — so the gate silently received
expected := [".olean"]
one entry with an empty name, and rejected the baseline via its own
fail-closed absence check. Exit code right, reason wrong.
· lift-guard.sh (new, pinned): derives every variable the lifted payload
reads, subtracts what the driver defines, and fails AT LIFT TIME naming
the remainder. Wired into all five lifting self-tests; selftest-shapes.sh
drops its inline copy. One implementation, not five, since drifting out
of sync is the entire failure mode.
· selftest-axgate.sh: lifts PROOFS verbatim, and its case 2 now poisons a
MANIFESTED leaf module rather than adding a stray file. The stray-file
attack was a no-op against membership semantics — that is the dead-file
gate's job, proven by selftest-harness case 8 — and it is also the
weaker attack, since adding files to Proofs/ must clear a gate that
editing an already-manifested module does not.
· Phase 2c's accounting identity moves behind its own marker with its own
ACCTFAIL verdict, so the phase is liftable by construction rather than
by a self-test knowing where to stop. Truncating the lift range alone
would have dropped Phase 2c's own verdict and left the test unable to fail.
· The other four lifting self-tests assert on the PAYLOAD rather than the
assembled driver, so a marker in the preamble cannot satisfy a check
meant to prove the lift landed.
Certified: 8/8 self-tests green in all four forks (~212 assertions, none
failing); check.sh bytes unchanged and still matching the pin written before
the 3h08m round-9 sweep, whose 36 GREEN button rows therefore stand.
--audit-only re-run against the new HARNESS.sha256 in every fork.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
THE DEFECT, as found. GPT-5.6 showed the textual classifier could report
PROVEN for a name Lean resolves to an axiom: a definition appearing only
inside a `/- -/` comment was read as real. Repairing that exposed something
larger and already live. Aeneas wraps long declarations:
axiom
curve25519_dalek.edwards.EdwardsPoint.Insts.CoreOpsArithNegEdwardsPoint.neg
The old scanner required keyword and name on one physical line, so it matched
nothing there and SILENTLY DROPPED the declaration — no MODEL row, no PROVEN
row, no failure. Nine to ten externals per fork had no row at all, and the
tier-A/B `neg` was missing from every committed table while the trust
documents claimed that class was machine-checked. Four pinned tables were
wrong, in four repositories, with the buttons green over them. No attacker
was required; Aeneas's own formatting did it.
The lesson is not "write a better regex". The scanner was FAIL-OPEN: input it
could not parse produced silence instead of a stop. A gate that drops what it
cannot read is worse than no gate, because the green covers a gap that is
invisible in the diff.
THE REPAIR, in three layers that each do only what they honestly can.
· DISCOVERY stays textual, because the template cannot be imported — it
declares the same names as the model and they would clash. It now strips
NESTED `/- -/` blocks, reads a name that wraps to a later line, tracks
namespace/section/end, and FAILS CLOSED: an unparseable keyword exits 2
naming file and line.
· Phase 0d keeps the cheap pre-compile comparison against the table.
· Phase 2d is new and authoritative. After compilation it generates a temp
Lean driver (the Phase 2b idiom, so no permanent module and no manifest
churn), imports every non-template gen module, and asks env.find? what
each external IS: kind and originating module. Model module -> MODEL;
non-axiom in a generated module -> PROVEN; an axiom outside the model, or
a name that does not resolve, is a failure. Lean's verdict must equal the
committed table's, and the answer count must equal the question count so
a truncated resolver cannot pass.
THE TABLES were regenerated and verified three ways before being committed:
an independently written second scanner agreed on 228/228 rows across the
four forks; all 47 changed rows carry file:line evidence; and the PROVEN row
was put to Lean directly — kind = def, module = CurveField.Funs.
dalek 54 -> 64 rows anza 42 -> 51 risc0 48 -> 57 betrusted 47 -> 56
Also corrected: `from_residual` was recorded EXTRA in every fork; the
template does demand it, so it is MODEL. anza carried a spurious `hash|EXTRA`
row from a mis-parse.
NEGATIVE-TESTED. With Phase 0d neutered so Phase 2d was the gate under test,
a PROVEN row edited to MODEL produces:
SEMANTIC DRIFT: ...neg — table says MODEL, Lean says PROVEN
selftest-correspondence.sh grew 4 cases -> 7: GPT's comment counterexample,
the wrapped declaration that was actually live, and a fail-closed check.
WHAT THIS DOES NOT CLOSE. Phase 2d resolves names in the environment built
from gen/. It establishes what a name IS — assumption or proof, and where —
not that the extraction faithfully represents the Rust. That remains the
Charon/Aeneas trust assumption and no gate here can close it.
Certified: both buttons, all four forks, purged trees. 8/8 green, with
62/48/53/52 externals resolved by Lean and every verdict matching.
THE GAP. The capability matrix in the control repo — built after the author
answered a capability question by grepping for a FILENAME rather than for the
property, three times in one session — asked whether each repository's auditor
enumerates itself. The four ed25519 forks answered no. Following that up found
something larger: `Proofs/Audit.lean`, the statement-binding driver from P1-a,
is compiled, is a member of the manifest, and was enumerated by NOTHING. The
kernel counted 3058 declarations where the inventory accounted for 3022, and
26 of the 36 missing were its.
THE PROPERTY, now enforced: every constant the kernel sees under this button's
manifest must appear, BY NAME, in either the corpus inventory or the
instruments' own surface. `emitDrivers` in InventoryCore walks the audit
modules and fails closed on an axiom (which would widen the trusted base
outside every cone) or on a standalone claim (which no certificate covers and
no allowlist pins), while admitting the obligations the elaborator generates
for a definition declared alongside it.
TWO WRONG FORMULATIONS FIRST, both recorded because the second is instructive:
· `kernel = corpus + instruments + N_DRIVERS`, where the last term was
justified as a per-driver "self-observation blind spot". It fitted dalek
and anza (2 drivers, residual 2) and broke on risc0 and betrusted
(1 driver, residual 2). The residual is 2 everywhere and has nothing to do
with drivers. This was curve-fitting from a sample of one, and it was named
as the highest-risk claim in the round-7 self-assessment BEFORE the data
refuted it.
· distinct-by-name counting, which collapses `CurveFieldProofs.zero_spec` —
a name that genuinely denotes two different declarations, in Proofs.Basic
and Proofs.ConstSpecs, walked by two drivers with separate environments.
THE MEASURED CAUSE of the residual. Lean materialises equation lemmas lazily,
when something forces an unfold, and each module that forces one gets its own
copy in its object file. `CurveFieldProofs.denote.eq_1` sits in both
SubNegSpec.olean and ConstSpecs.olean; `CurveFieldProofs.limbsVal.eq_1` in both
ReduceSpec.olean and ConstSpecs.olean. The kernel gate reads object files and
counts both copies; the environment holds one constant per name. Counts cannot
relate those two views in either direction — so the check compares SETS, which
is the idiom the rest of this estate already uses, and no constant remains that
could be widened to make a red run green.
Negative-tested: with the instrument walk disabled, the check names
Proofs.Audit's declarations as unaccounted.
ANZA also lacked one allowlist row, `subtle.Choice...from.eq_1`, the same
lazy-materialisation effect seen from the other side. Verified rather than
assumed before adding it: the parent is declared in gen/CurveField/
FunsExternal.lean — the model, pinned by bytes and not inventoried by
declaration — the lemma materialised in Proofs.CompressSpec which forced the
unfold, and all three other forks carry the lemma row while none carries the
parent.
CONSEQUENCE WORTH DISCLOSING: the audited declaration surface is not purely a
function of the corpus source. It depends on which proofs forced which
unfolds. This is now a known gap in the round-7 kit.
Also corrected here: Proofs/Inventory.lean's header claimed the audit drivers
were excluded from the compile manifest. They are lines 42-44 of PROOFS.
Certified by a full sweep: both buttons, all four forks, purged trees. 8/8.
Aeneas emits a *_Template.lean naming everything the extracted code needs
from outside itself — the extraction's own statement of its boundary.
extract.sh has always said, in prose, "after regenerating, diff the template
against the hand-written file". Prose is not a gate, and the diff cannot be
one: the two files legitimately differ in almost every line, holes and
Aeneas comments against real definitions and modeling policy.
MEASURING FIRST CHANGED WHAT THIS ITEM SHOULD BE. The TODO offered two
options — enforce the diff, or pin both files — and the answer turned out to
be neither. Both files were ALREADY byte-pinned by Phase 0b. And two further
things stand here: the generated Funs.lean imports the model and CALLS these
externals, so the Lean compiler enforces their TYPES wherever the extracted
code uses them; and the per-certificate exact cones catch any external that
becomes, or stops being, an assumption anything depends on.
What none of those three sees is the CLASSIFICATION: for each name the
extraction asks for, whether this repository answers with an ASSUMPTION or
with a PROOF. That is the tier-A/B claim the documents make in prose — the
curve calls and the three curve types resolve to proven definitions rather
than axioms, because gen/CurveField/Funs.lean opens `namespace
curve25519_dalek` and so defines the very names Aeneas asks for. Nothing
checked it. A regeneration that renamed one, or a model that quietly
answered one with an axiom instead, would have left the documents claiming a
proof where the repository had an assumption.
Phase 0d recomputes the classification with model-correspondence.py
(namespace-aware, so a definition inside a namespace counts under its full
name) and requires equality with the committed MODEL-CORRESPONDENCE.txt.
UNRESOLVED — the extraction asking for something nothing here provides — is
a hard failure.
dalek 43 MODEL 8 PROVEN 3 EXTRA
anza 38 MODEL 0 PROVEN 4 EXTRA (no CurveSig crate)
risc0 36 MODEL 8 PROVEN 4 EXTRA
betrusted 35 MODEL 8 PROVEN 4 EXTRA
selftest-correspondence.sh, five cases, negative-tested by disabling the
comparison. The case that matters is 2: a PROVEN external answered by an
axiom instead. No name changes anywhere, every byte pin still matches, and
it compiles, because the signature is unchanged — before Phase 0d nothing in
the button could tell.
Trap recorded for whoever extends it: case 3 first deleted the PROVEN rows,
which was VACUOUS on anza, since anza has none — it removed nothing, the
table still matched, and the case passed while testing nothing. It now
deletes the first row whatever its verdict AND asserts the file changed.
extract.sh now points at the gate instead of asking a human to look.
Certified by a full sweep: both buttons, all four forks, purged trees,
machine otherwise idle. 8/8 green.
Phase 2c exists because a source-regex enumerator proved evadable in
ltl-accumulator-verified: attributed, private and `instance` declarations
and a nested-namespace basename collision all slipped past it. The fix was
to stop reading source text and ask the Lean environment, and that fix was
ported here. But a fix ported is not a fix tested. selftest-inventory.sh
proves the GATE reacts to a difference; it feeds synthetic observations and
never runs the walker. Nothing here had ever asked whether the WALKER SEES
a declaration written in an evasive shape.
selftest-shapes.sh adds all four shapes to an audited module, recompiles it,
runs the real Phase 2c, and requires each one to be NAMED in the
UNCLASSIFIED list. Asserting that the gate merely failed would not do: one
shape surfacing fails the run while the other three ride along unseen. All
four forks report all four. Negative-tested by removing the injection — the
run then reports the walker blind and fails.
The victim module is derived from each repo's own manifest, not named: the
forks do not share a corpus (dalek/anza attack Proofs.Basic, risc0/betrusted
Proofs.DecompressMain), and a hard-coded name would have silently found
nothing on half of them. It must be manifested, must not be an inventory or
audit driver, and must be imported by no other manifest module.
Two notes for whoever edits this next. When re-deriving a leaf module, the
inventory drivers must be excluded from the set of IMPORTERS as well as from
the candidates: they import the whole corpus, so leaving them in makes every
module look imported, finds no leaf, and the test silently has no victim at
all. And a lift of Phase 2c needs SCALAR_SH/SCALAR_MANIFEST alongside
PROOFS, or the coverage check dies on an unbound variable.
New executable pinned in HARNESS.sha256. Certified by a full sweep: both
buttons, all four forks, purged trees, machine otherwise idle. 8/8 green.
selftest-tiers.sh tests the property this repository exists to assert and
that nothing had tested: the arithmetic tier rests on the three kernel
axioms and NOTHING else. Five cases, all green.
control both tiers pass on the untouched tree
case 1 an apex axiom injected into an ARITHMETIC certificate's proof,
statement untouched so only the cone moves -> Phase 3 rejects
case 2 apex boundary widened by one name -> Phase 3 rejects
case 3 apex boundary narrowed by one name -> Phase 3 rejects
restored both tiers pass again
Which axiom to inject is derived per fork from this repo's own documented
boundary intersected with the victim module's import closure; no name is
hard-coded, so the same script ships unchanged in all four forks.
Two defects in the test were found and fixed before it was trusted. The
lifted driver first omitted `set -euo pipefail`: the phase's Lean work runs
in a subshell and the phase ends in a bare `echo ""`, so without -e the
subshell's exit 1 was masked and the driver reported GREEN while printing
APEX AUDIT FAILED. And a line-count sanity check passed an empty driver
because the CERTS array padded it; the guard now looks for the diagnostics
it means to provoke. Both are recorded in the script's comments.
The test restores what it touches and rebuilds the module it edits, so it
leaves the tree exactly as it found it. New executable is pinned in
HARNESS.sha256 (Phase 0c fails closed on an unpinned one).
Certified by a full sweep: both buttons, all four forks, from purged trees,
machine otherwise idle. 8/8 green, 0 errors.
Phase 0a purges every .olean before compiling, bans stray Lean files at the
verification root (LEAN_PATH contains $PWD, so they join the build unaudited),
and requires gen/ to be exactly the model manifest plus its pinned templates.
The templates are KEPT, unlike SLH-DSA which deletes them: extract.sh directs
the operator to diff the hand-written external models against them, so they are
the reference for that comparison and P2-c will enforce it.
The purge is skipped under --audit-only, which exists to audit the artifacts a
previous full run produced. Those two features would otherwise destroy each
other, and it is a further reason an audit-only transcript is not evidence: it
has not had this hygiene applied.
WHAT THE PURGE EXPOSED, and it is the point of the whole item:
This button had never compiled the corpus from nothing. The signature apex
rests on scalar arithmetic — PointLiftSpec -> ScalarPackSpec ->
ScalarFromBytesSpec, and SigApexSpec -> ScalarDenote — and TWELVE of the scalar
layer's thirteen modules are transitive prerequisites of this manifest. They
were never compiled here. The button worked because check-scalar.sh had run at
some earlier point and left its .olean files behind. .olean is gitignored, so
no git status could ever have shown that the verdict rested on untracked
artifacts produced by a different script.
Nothing about the proofs was wrong. The evidence was resting on something
invisible, for the entire life of these repositories, and it surfaced the
moment something finally cleaned up before verifying.
Those twelve are now compiled here as PREREQ — BORROWED, NOT OWNED.
check-scalar.sh still audits them; Phase 1b asserts every borrowed name belongs
to the other manifest and to neither twice, so the list cannot become a second
ownership claim.
Two consequences fixed along the way, both the spelling-versus-membership error
that ScalarPackSpec has now taught four times:
- Phase 2b globbed Proofs/*.olean and would have demanded artifacts this
button never builds. It now scans its manifest by membership and fails
closed on a missing one.
- The three inventory drivers were exempted from the dead-file gate and
compiled in a later phase; after a purge they were absent when Phase 2b
ran. They are now in the manifest like everything else, and three
exemptions are gone.
The sweep runner now reports RESOURCE rather than RED when it sees a
memory_exception: lean-guard's clamp is not a broken proof, and it has misled
the operator once and the author once.
Verified green: 8 full runs from completely purged trees — four check.sh, four
check-scalar.sh — zero red, zero resource. Every artifact rebuilt from
committed source. These are the first runs in this repository's history whose
verdict provably depends on nothing but the bytes in git.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Gate work dominates this estate's wall-clock: on 2026-07-29, 3.9 hours of a
session went to Lean re-elaborating proofs nobody had edited while the audit
phases themselves took about fifteen seconds. --audit-only runs every gate
against the artifacts a previous full run left behind: ~60s against ~1280s.
IT IS SAFE ONLY BECAUSE IT REFUSES.
- It requires every shipped .lean to be BYTE-IDENTICAL to a basis recorded by
a previous full run. Not mtimes: `touch` defeats those, and a stale-artifact
check that fails open is worse than no shortcut at all, because a green
button would then describe a corpus that is no longer on disk.
- The basis is gitignored build state, so a fresh clone cannot inherit
permission to skip compiling.
- The closing banner differs and says in words that the run is not evidence.
selftest-auditonly.sh exercises seven cases: no basis, an edited comment
character, a deleted source, a new source, a missing artifact, a truncated
basis, and — asserted as a PASS — every source's mtime touched with bytes
unchanged, which pins the bytes-not-mtimes decision rather than leaving it
implicit. Negative-tested: with the basis comparison disabled a changed source
is wrongly accepted, exit 0 and zero refusals, so the guard is load-bearing.
A PHASE TERMINATOR, because this broke twice. Every self-test lifts a phase
from check.sh by scanning to the next phase marker. The last phase had no
marker after it, so a lift ran to end-of-file and swallowed whatever was
appended later — first Phase 2c into the axgate lift, then T1's tail into the
binding lift, where it referenced $AUDIT_ONLY and died under `set -u`. Both
surfaced as the BASELINE case failing: a self-test blaming a gate for its own
extraction bug. The phases now end at an explicit sentinel and both lifters
stop there, so nothing appended below can silently become part of the last
phase from a lifter's point of view.
TRUSTED-BASE.md records what an audit-only transcript does and does not
establish, and — because it cost a confusing red run today — that lean-guard's
memory clamp presents as `FAIL: Proofs/<module>` while being a resource
condition, not a broken proof.
Verified green: 8 full button runs (four check.sh, four check-scalar.sh) and 20
self-tests across the four repositories, zero red. One earlier run failed on
the memory clamp because the author ran a test suite concurrently; re-run on a
quiet machine, green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
THE SEAM. This repository is checked by two scripts, and until now neither
asserted anything about the other's scope. check.sh's dead-file gate simply
SKIPPED anything named Scalar*, so a new Proofs/ScalarX.lean was gated by
nothing at all: absent from one manifest by exemption, from the other by
omission, compiled by neither, inventoried by neither. Each button now reads
the other's manifest and requires every shipped proof source to belong to
EXACTLY ONE of them — neither orphaned nor double-claimed, both directions,
plus a phantom check on entries naming files that do not exist. Negative-tested
four ways, including the exact hole this item names.
THE SCALAR BUTTON. Closing the seam exposed it as the estate's weakest link,
having been left behind by every hardening round while the main button gained
five phases. 45 lines to 227:
- source-integrity check over its sources;
- harness-pin verification, so running THIS button alone is protected and not
only running it after check.sh;
- a kernel-side axiom-declaration gate over the compiled artifacts, replacing
a source-text grep that is evadable four ways on v4.30.0-rc2;
- a declaration inventory of ~1880 constants against its own allowlist,
diffed both directions with a count trailer. These 13 modules were the only
part of the proof corpus with no inventory: check.sh Phase 2c named them as
uncovered on every run, and now names the button that covers them instead;
- per-certificate exact-cone assertions replacing `-eq 13` over matching
output lines. A count cannot say WHICH certificate is clean and passes just
as happily if one cone is reported twice.
Every fork-specific fact was read from the existing script rather than assumed:
risc0 and betrusted audit sub_loop1_one_spec where dalek and anza audit
cond_add_l_one_spec, untouched.
THREE BUGS, ONE ROOT CAUSE, all found by the gates rather than by review. Each
reasoned about how a thing is SPELLED instead of what it BELONGS TO, and the
corpus punished each: Proofs/ScalarPackSpec.lean is named like the scalar layer
and owned by the main button.
- the scalar dead-file gate globbed Scalar* and demanded ScalarPackSpec be
scalar-owned. REMOVED rather than special-cased: the seam check tests
membership in exactly one manifest, which is strictly stronger than any
prefix;
- the scalar axiom gate scanned Scalar*.olean, reporting "14 modules" for a
13-module manifest. On a tree where check.sh had not run that artifact is
absent and the button would have failed for a false reason. It now scans
the manifest by membership and fails closed on a missing artifact;
- Phase 2c's driver discovery globbed Inventory*.lean and claimed the other
button's driver, then correctly complained its own manifest lacked those
modules.
This is the family the campaign began with: a source-text axiom grep reasoning
about spelling. Recorded in TRUSTED-BASE.md because it generalises.
Also fixed: the first negative test of the scalar gate's absence check passed
for the wrong reason — the button recompiles before the gate runs, so removing
an artifact merely caused it to be rebuilt. Retested against the lifted phase,
where absence is a persistent condition.
Verified green: 24 runs across the four repositories — four main buttons, four
scalar buttons, and sixteen self-tests — zero red.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 2b asks the kernel whether any AXIOM is declared under Proofs/. Phase 3
pins the cones of the named certificates. Between them sat every other
declaration in the corpus — around three thousand of them — and a helper lemma
quietly acquiring a hash oracle in its cone moved nothing either phase looked
at.
Phase 2c closes that. Ported from ltl-accumulator-verified, where a nine-attack
self-test proved a source-regex enumerator evadable by attributed, private,
indented and `instance` declarations and by a nested-namespace basename
collision. Reading the compiled environment sees what the kernel saw; no name
shape hides. Every constant contributes module, name, kind and full axiom cone,
and the observed set must equal inventory-allowlist.txt exactly in BOTH
directions, with a count trailer so a truncated run cannot pass as an empty
diff.
FOUR THINGS THIS BUILD GOT WRONG, each caught by a check rather than by review:
- The number of inventory drivers is a per-repo FACT, not an assumption.
dalek and anza cannot import their corpus as one environment (Proofs.Basic
and Proofs.ConstSpecs both declare CurveFieldProofs.zero_spec); risc0 and
betrusted have no Proofs.Basic at all. Determined by compiling a probe.
check.sh now DISCOVERS its drivers from the filesystem instead of naming
two, and the generator refuses to split out a module the repo lacks.
- The split let one real declaration hide behind another's entry. Keyed on
name alone, the two zero_specs produced byte-identical records, so 3022
declarations were covered by 3021 allowlist entries. Caught by the count
trailer. Every record now carries its originating module.
- The gate's success line said "single sanctioned axiom", inherited from the
accumulator's policy. This corpus permits NONE. A success message
describing a different rule is how an assertion stops meaning anything.
- selftest-axgate.sh lifted Phase 2b with a range ending at "Phase 3", so
inserting Phase 2c between them made it swallow the new phase and die on
variables only check.sh defines — surfacing as the BASELINE case failing,
a self-test blaming a gate for its own extraction bug. Both self-tests now
stop at the next phase marker whatever it is called, and refuse to run if
they capture more than one phase. The guard is the fix; the range was the
symptom.
WHAT THIS IS NOT, recorded in TRUSTED-BASE.md at the same length as the claim:
- No independent cone walker. The accumulator cross-checks collectAxioms
against a hand-written walker. Ported here it was wrong in BOTH directions
on mathlib's inductive shapes: EdPoint gave [] against the kernel's three
axioms, and once extended, ProjPoint gave three against the kernel's none.
Two implementations disagreeing both ways are a second wrong answer, not a
check. These cones rest on collectAxioms alone.
- Thirteen Proofs/Scalar* modules are inventoried by nothing — the
second-button seam, still open. Phase 2c names every uncovered module on
every run so the omission is visible rather than inferred.
selftest-inventory.sh exercises the shipping gate with six cases, each
asserting a specific diagnostic, including the one that matters: a cone
widened by one oracle while name, module and kind stay put. Negative-tested by
disabling the gate's diff, which turns two cases red including one for the
wrong reason, correctly reported as such.
Verified green: 20 runs across the four repositories (four buttons, four
harness, four inventory, four axgate, four binding self-tests), zero red. The
four check-scalar.sh greens from the preceding sweep stand: that script neither
reads the pin file nor changed.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Every gate this repository has was executed by scripts that nothing pinned.
Round-5 review of the companion SLH-DSA repository stubbed the compiler
wrapper alone and its button printed ALL GREEN in 3.6 seconds over
deliberately destroyed proofs; flipping two guards in the audit driver
disabled every check with the digest byte-identical. Depth of checking is
worth nothing if the thing doing the checking is unbound — and every gate
added this week made that gap more valuable to an attacker, not less.
Phase 0c requires every harness file to match HARNESS.sha256. Two design
points carry the weight:
- WHICH files must be pinned is POLICY and lives in check.sh, never in the
map being consulted. If the required set were read from the pin file,
deleting an entry would silently un-pin that file. It is instead derived
from the filesystem, so a deleted entry is a set mismatch and a build
failure. That is the exact defect SLH-DSA round-6 found, closed here by
construction.
- Membership self-derives from the executable bit: anything this script can
shell out to must be pinned, so a NEW script fails closed until someone
pins it deliberately. Load-bearing files that are not executable — the
audit driver, the committed manifests, the policy tables — cannot be
discovered that way and are listed explicitly.
lean-guard is inside the set, which finally makes the standing "lean-guard
stays hash-pinned" rule a property of the repository rather than a convention.
selftest-harness.sh replays five cases, each asserting a specific diagnostic:
an edited lean-guard, a new unpinned executable, a deleted pin entry, a
missing pin file, and a positive control. It was itself negative-tested — with
the hash comparison removed it goes red on exactly that case while cheerfully
reporting "10 harness files match their pins".
TRUSTED-BASE.md states the limit at equal length to the claim: pinning a
harness from inside that harness is circular, and an author who edits a script
and refreshes its pin in the same commit passes every phase. What the pin
changes is that the edit can no longer be SILENT — it must appear in the diff
at the commit being reviewed. A green button says "this is the apparatus that
was reviewed", never "this apparatus is trustworthy".
Also fixed, found by this sweep: both self-tests compared the working tree
against its starting state with `diff <(echo "$VAR") <(command)`, which is
asymmetric — for a clean tree the variable is empty and `echo` emits a blank
line the command does not. It reported a difference precisely when nothing was
wrong, and only surfaced once P1-a was committed and Proofs/ became clean.
Both now compare as strings.
Verified green: 20 runs across the four ed25519 repositories (four buttons,
four harness self-tests, four axiom-gate self-tests, four binding self-tests,
four scalar buttons), zero red.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phases 3/3b establish what each certificate RESTS ON. Neither says what it
SAYS, nor what it is ABOUT. A certificate gutted to a tautology of the same
axiom cone passes both; so does one whose reference definition has been
redefined to BE the extracted code, at which point the theorem reads
`loop = loop` and every cone is byte-identical.
Phase 3c closes that. Proofs/Audit.lean emits a canonical block holding the
policy constants, every certificate's fully-elaborated statement (pp.all, so
implicit arguments, instances and universe levels are visible), and the body
of every specification constant transitively reachable from those statements.
Its SHA-256 is pinned in check.sh and the block itself is committed as
AUDIT-MANIFEST.txt, so a mismatch is DIFFED, not merely reported. 31
certificates, 68 specification constants per repository.
Two tiers, not one. These forks have an arithmetic tier that must stay
oracle-free and an apex tier carrying this fork's hash and wire-format axioms,
and the apex boundary genuinely differs per fork (dalek 8 extra names, anza 4,
risc0 and betrusted 5). One shared constant would have widened the arithmetic
tier to accept hash oracles, which is the most valuable property these repos
have. Each auditor is generated from its own repository's policy.
Phase 0b pins the extracted model. This was not a precaution: risc0 and
betrusted were observed emitting BYTE-IDENTICAL audit-manifest digests
(6c821b8e…) while shipping demonstrably different extracted models, their
point-doubling routines differing in operation order. A statement names an
extracted function; it does not contain that function's body. Binding
statements is not binding the subject. Membership derives from the filesystem,
so a new model file fails closed.
selftest-statements.sh attacks both phases with ten cases, each asserting a
specific diagnostic: an edited model body, an unlisted model file, a widened
policy, a hand-edited committed block, a certificate dropped from the auditor
WITH the digest refreshed to match, and a gutted statement whose cone is
unchanged. It lifts the phases out of check.sh at run time, so it attacks the
shipping gate rather than a copy.
Two bugs found and fixed during that testing, both mine: Phase 3c read `$0`
after `cd "$AENEAS_LEAN"`, and $0 is the caller's relative path; and the
axgate self-test compared the tree against a pristine checkout rather than
against how it found it. A third expectation was wrong rather than the code —
widening the apex boundary is caught by the exact-cone requirement before the
digest ever runs, which is a stronger rejection, and the test now says so.
All sixteen runs green at these commits: four main buttons, four axgate
self-tests, four binding self-tests, four scalar buttons.
TRUSTED-BASE.md records what this binds and, at equal length, what it does
not: a digest binds identity, not meaning; an author can rotate the pins in
one commit and is caught by review, not by the script; and pinning the model
says nothing about whether Charon and Aeneas translated the Rust faithfully.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds one item naming what check.sh Phase 2b binds — every compiled
Proofs/*.olean, kernel-side, membership self-derived, fail-closed on a
missing module — and, more importantly, what it still does not bind:
declarations, not statements. A theorem gutted to a tautology with the same
axiom cone passes every phase. Reading the statements remains a human act,
and this document is where that has to be said rather than left for a
reviewer to discover.
Verified green at this commit's parent across all eight buttons on
2026-07-28; see formal-verification-control/RECORDED-RUN-2026-07-28.md.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 1's anti-smuggling check reads source text. Measured today on Lean
v4.30.0-rc2, four distinct declarations compile cleanly and slip past its
anchored pattern:
` axiom cheat : ...` one leading space
`@[simp] axiom cheat : ...` line starts with the attribute
`unsafe axiom cheat : ...` `unsafe` absent from the modifier list
`axiom` <newline> ` cheat` no space follows the keyword
Any of them yields a repository that proves False while the button prints
ALL GREEN. Only the tab variant is blocked, and by Lean, not by us.
Hardening the pattern would fix the exhibited syntax rather than the class,
which is the mistake this estate has made before. Phase 2b stops parsing text
and asks the kernel instead: it reads every compiled Proofs/*.olean with
readModuleData and rejects any declaration that is an axiom.
Design notes:
- reads compiled artifacts rather than importing the modules, because
Proofs.Basic and Proofs.ConstSpecs deliberately reuse `zero_spec` and a
whole-corpus import is impossible by construction;
- membership is self-deriving from the filesystem, so Scalar* and
AxiomCheck are covered too — both are skipped by the CERTS audit and by
the dead-file gate;
- fails closed on absence: a missing .olean would make the scan vacuous, so
the count of compiled modules must equal the count of shipped sources;
- removes its temp source AND artifact on both paths, since a bare `rm`
after the call never runs under `set -e` when the gate goes red — exactly
how this repo accumulated 101 orphan .olean files;
- ~3 s for the whole corpus, against ~53 s for one module-importing run.
Phase 1's grep stays as a fast first line of defence. Phase 2b is the gate
that is load-bearing.
selftest-axgate.sh attacks the shipping gate, lifted out of check.sh at run
time rather than copied. It asserts the specific diagnostic, so a rejection
for an unrelated reason fails too, and it was itself negative-tested: with
the gate's throwError removed, the self-test goes red on exactly that case.
No proof, statement, specification or certificate is touched. No attested
commit is altered — the log binds specific commit hashes, all of which remain
ancestors of HEAD.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Phase 3 and 3b mktemp an audit file, compile it, then removed only the .lean —
leaving the .olean behind on every run. check-scalar.sh:38 has always done this
correctly (`rm -f "$AUD" "${AUD%.lean}.olean"`); the main script was the odd one
out. Estate-wide that had accumulated 101 orphan compiled modules with no
sibling source (dalek 44, anza/risc0/betrusted 19 each), invisible to git
because *.olean is gitignored. All swept.
The litter was inert — the names are not valid Lean identifiers, so nothing
could import them. It matters as a pattern: an audit whose verdict can depend on
untracked build state is the class of defect that took eight review rounds to
close in the sibling SLH-DSA repository (there: an orphan .olean with its source
deleted satisfied an import and the button went green). A build-hygiene phase
that purges compiled artifacts and bans stray files is the proper fix and is
queued as part of the protocol port; this commit stops the bleeding.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- README: pyramid-diagram apex row upgraded to the proven full lift
(accepted <=> decompress(R) = [k](-A)+[s]B), status table names all
four button-enforced tiers, apex section gains the phase-2 tier table
(half-lift / point equation / full lift) + the decompress-chain
summary; source pin updated to the pushed patch commit.
- TRUSTED-BASE item 5: rewritten from the single byte-apex certificate
to the FOUR enforced tiers (decompress_of_canonical noted as
standard-three-only).
- gen/CurveField/FunsExternal.lean: stale root-namespace
edwards.decompress.step_1/step_2 axioms removed (dead weight left
behind by un-opaquing; outside every cone, but they forced
fully-qualified unfolds - see control FAILURES.md).
- check.sh Phase 3b success echo aligned to "apex + full-lift" (echo
only; the enforcing greps covered all four tiers already).
Validated by the pass-4 sweep: 9/9 buttons green (this repo's check.sh
+ check-scalar.sh among them), logs retained in the pass workspace.
Full record: formal-verification-control/COHERENCE-PASS-4.md.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
verify_accepts_iff_point_eq, button-enforced
Port of dalek's PointEqSpec (the encoding-injectivity mathematics is
fork-independent; compiled first try): for any valid on-curve point Q
whose canonical encoding is the signature's R bytes, the verifier
accepts IFF Q equals the recomputed point as denoted affine points -
the literal point-level EdDSA verification equation, no decompress
needed. enc_point_inj carries the standard three axioms; the equation
itself carries exactly this fork's enforced apex boundary, and Phase 3b
now audits all three tiers (byte apex, half-lift, point equation).
Full button green fresh.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
THE HALF-LIFT IS NOW PROVEN ON ALL FOUR PYRAMIDS. anza's shape: the hash
oracle is one sha512_hash3 bind, -A is the STORED minus_A field (no
negation call), and the scalar arrives already parsed - so the recompute
inversion is five flat bind_ok_inv steps (two anonymous slice reads, the
oracle, the reduction, the dsm) and the assembly takes ExtValid/OnCurveExt
of self.minus_A directly. Files 1-4 are dalek's verbatim modulo the
curve25519 namespace; every proof compiled FIRST TRY.
verify_accepts_iff_point: accept IFF the signature's R bytes are the
canonical encoding of the recomputed [k]*minus_A + [s]*B (valid,
on-curve, certified model). Five new standard certificates; Phase 3b
enforces anza's tight boundary (SHA-512 oracle + foreign Signature type
+ its two accessors) on BOTH apex and half-lift. Full button green fresh.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- README: the pyramid diagram claimed the cofactored ZIP-215 equation,
which is NOT the proven statement - corrected to the actual theorem
(accepted IFF compress([s]B-[k]A) = R, byte-for-byte) and the signature
row now names verify_accepts_iff; new "The signature apex (phase 1)"
section states the theorem, this repo's glue architecture, the exact
button-enforced axiom cone, and the phase-2 deferral.
- TRUSTED-BASE: item 5 rewritten from an aspirational hash paragraph to
the structural boundary - certificate name, exact allowed cone, and the
Phase 3b enforcement that fails the build on any deviation.
- Dead pre-merge artifacts removed: gen/CurveScalar, CurveScalar.llbc,
extract-scalar.sh (the merged gen/CurveField universe is the single
model; check-scalar.sh remains the scalar button, header updated).
- lean-guard: Guard 3a retry ladder (LEAN_MEM_WAIT_SEC) - a clamped run
that dies on memory retries as headroom improves, converting ambient
memory pressure from a deterministic abort into a delayed pass.
Fresh green buttons after these changes: check.sh (incl. Phase 3b apex
audit) + check-scalar.sh, both at shipped defaults, coherence pass 3
sweep 2026-07-05.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
FOURTH AND FINAL PYRAMID CAPPED - the signature layer is complete on all
four ed25519 forks. anza's verify code lives in the same crate as the
curve (solana-ed25519), so the whole verify path joins the merged
CurveField extraction directly: one universe, no glue layer, no FQ-name
welding, and the Error enum plus the parse/filter helpers are all real
extracted code.
- extract.sh: verify_sha512 start-from joins the merged stanza;
sha512_hash3 and the foreign ed25519 crate opaque; RUSTFLAGS
--cfg curve25519_serial_only pins the serial backend so
get_selected_backend extracts as the real constant Serial (the stale
dispatch axiom is deleted from FunsExternal).
- gen/CurveField externals: real defs for the ?-operator plumbing
(Try::branch, FromResidual) and faithful identity models for
Choice::unwrap_u8 (transparent-u8 body: self.0) and the RangeFull
get_unchecked[_mut] raw-pointer pair (Rust body returns the pointer
unchanged) - the three would-be cone intruders, eliminated.
- Proofs/SigApexSpec.lean: verify_loop_full (standard three-axiom cone)
and verify_accepts_iff - the verifier accepts IFF the recomputed
compress([k](-A) + [s]B) equals the signature's R byte-for-byte, with
the ZIP-215 legacy filters and the s < l parse conditioned by
hypotheses, mirroring the siblings' hparse.
- check.sh Phase 3b enforces the apex cone to be EXACTLY
[propext, Classical.choice, Quot.sound, ed25519.Signature,
ed_sigs.sha512_hash3, ed25519.Signature.r_bytes,
ed25519.Signature.s_bytes]
- the tightest boundary of the four pyramids: the SHA-512 oracle plus
the foreign wire-format type and its two byte accessors, nothing else.
check.sh (incl. Phase 3b) + check-scalar.sh both green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Same architecture as the dalek/risc0/betrusted forks: the Scalar52
arithmetic start-froms (11 fns) plus scalar::from_bytes_mod_order[_wide]
join the CurveField extraction, so the field, curve, and scalar layers
share a single type universe - the prerequisite for the signature apex,
whose verify glue must see curve AND scalar calls resolve to proven
definitions by fully-qualified name.
Proofs/ScalarDenote.lean flips its import CurveScalar.Funs ->
CurveField.Funs (one line; the whole scalar proof chain recompiles
unchanged on the merged gen). check-scalar.sh repoints its GEN list.
gen/CurveScalar retained until the deprecation pass, as on the siblings.
check.sh + check-scalar.sh both green, all certificates axiom-clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The complete non_adjacent_form(5) verification (four stages):
- `Proofs/DsmNafLoadSpec.lean` (generated) — the LE byte-to-word load.
- `Proofs/DsmNafMath.lean` — the digit loop's arithmetic core: window-read
lemmas (single/cross-word), the exact ZZ invariant steps (Nat.mod_mul
telescope), the carry-kill argument from V < 2^253, and the exit theorem.
- `Proofs/DsmNafLoopSpec.lean` — the w=5 digit loop by induction on the
remaining-bits measure: per-step 64-bit window read (4-way word split),
digit write via hcast/wrapping_sub (exact value window - 32*carry',
oddness, |d| < 16), invariant carried through even/odd steps.
- `Proofs/DsmNafSpec.lean` — the public spec: both entry masserts
DISCHARGED; the digits satisfy the NAF conditions and
sum naf[k]*2^k = V EXACTLY (integers, no modular slack)
for any scalar whose LE byte value V is below 2^253.
And the campaign's brick 4, `Proofs/DsmMulSpec.lean`:
- `run_basepoint` — the transpiled ED25519_BASEPOINT_POINT is the standard
base point: valid extended coordinates (X*Y = Z*T) and the curve equation,
kernel-checked via denominator-free 121666-scaled witnesses. Includes the
generic witness lemmas fp_mul_eq_of_witness / onCurve_of_witness.
- `vartime_double_base_mul_spec` — THE PHASE-1 COMPUTATIONAL SPEC of
vartime_double_base::mul: for canonical scalars and a valid on-curve A,
the result is valid, on-curve, and denotes
dsmFold (naf a) (naf b) (edPt A) edBasePt edId 256
with both digit arrays proven exact NAF encodings. Phase 2 (group
semantics [a]A + [b]B) requires Edwards associativity — deferred and
documented; nothing assumes it.
Also: removed a vestigial pre-re-extraction axiom stub
(backend.serial.scalar_mul.vartime_double_base.mul) from FunsExternal —
a root-level leftover that shadowed the real namespaced definition during
name resolution in proof files. Never referenced by any certificate (the
#print-axioms audit guards against that); deleted for hygiene.
CERTS += naf_load_spec, naf_exit, naf_digit_loop_spec,
non_adjacent_form_spec, run_basepoint, vartime_double_base_mul_spec —
each audited to exactly [propext, Classical.choice, Quot.sound].
Full check.sh green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Three new proof files over the CurveField extraction, composing the proven
group-law layer (no new axioms, no associativity assumed — computational
layering over the abstract `edAdd`):
- `Proofs/DsmTableSpec.lean` — `NafLookupTable5::from(&A)`: the 8 entries
are valid `ProjectiveNielsPoint` caches of valid on-curve points denoting
the odd multiples A, 3A, ..., 15A as the `edOdd` double-and-add recursion.
7 explicit loop peels over edwards_as_projective_niels_spec /
add_projniels_law / compl_as_extended_law, seeded by edwards_double_law.
`select`: both masserts (x odd, x < 16) DISCHARGED — panic-freedom is
proven, not assumed; post enumerates all 8 digit cases.
- `Proofs/DsmStepSpec.lean` — `proj_double_law` (the projective doubling
denotes `edAdd P P`; same Z^2-scaled linear_combination discipline as the
extended-coordinate law), `compl_as_projective_law` ((X:Z),(Y:T) to
(XT:YZ:ZT) preserves the point), `naf_select_entry` (digit-indexed lookup
returns THE entry: NafEntryOf r A ((x-1)/2)), and `dsm_step_p_law` /
`dsm_step_b_law`: the three-way NAF digit step denotes `edDigit` — add
the d-th odd multiple, add its negation, or pass through.
- `Proofs/DsmLoopSpec.lean` — the 256-iteration Straus loop by GENUINE
induction on the counter (one symbolic body walk, no unrolling):
`dsm_loop_spec` — from the identity, the loop returns a valid on-curve
point denoting `dsmFold ... edId 256`, the abstract double-and-add fold
of both digit arrays over the table points. Digit and table hypotheses
are exactly what the NAF spec and naf_table_spec provide (layering).
check.sh wired: PROOFS + AUDIT_IMPORTS + 7 new CERTS (naf_table_spec,
naf_select_spec, proj_double_law, compl_as_projective_law, dsm_step_p_law,
dsm_step_b_law, dsm_loop_spec), each `#print axioms`-audited to exactly
[propext, Classical.choice, Quot.sound]. Full check.sh green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
extract.sh now opens crate::backend::serial::scalar_mul::vartime_double_base
(the other scalar_mul strategies stay opaque): non_adjacent_form (with its
loops), NafLookupTable5 (from/select), the curve-model helpers and
vartime_double_base::mul itself land in gen/CurveField - the same
namespace as the proven edwards operations, so the coming double-and-add
induction can consume EdDouble/EdAddProjNiels/EdConvert directly.
Zero sorries, zero external axioms (the pinned sources carry documented
compat refactors: single-assignment loop helpers, param-rooted while,
always-256-iterations, index-based LE load).
Full check.sh pressed fresh over the regenerated model: every existing
field and group-law certificate still green and axiom-clean - the scope
extension is purely additive.
The apex brick of the scalar layer: for any 64 bytes (the opaque SHA-512
digest), [from_bytes_wide bytes] = (LE 512-bit value) mod l, with
canonical 52-bit-bounded output. Composition: bytes_unpack_spec (8x8
loops) -> split_words_lo/hi_spec (exact div/mod per limb, disjoint ORs
as additions) -> wide_split_telescope (isolated omega) -> montgomery_mul
by R and RR (R cancels as a unit, RR restores it) -> the canonical add.
The two kernel-capacity walls found and crossed en route (control repo
FAILURES.md updated):
- a montgomery_mul inside any walk motive replays its 400-line body at
every kernel step (fix: named prefix functions in the pinned source);
- straight-line IndexMut closure chains make kernel defeq exponential in
depth (fix: struct-literal construction - the split halves now build
Scalar52([...]) directly). Full certificate: 77 s kernel-inclusive.
Regenerated gen (sources factor from_bytes_wide -> from_bytes_wide_parts
-> split_words_lo/hi; documented pure refactors, cargo-checked).
check-scalar.sh: 13 proof files, 13 kernel audits, all exactly
[propext, Classical.choice, Quot.sound]. Button pressed fresh: green.
Toward Scalar::from_hash: bytes_unpack_spec proves the from_bytes_wide
word-unpack loops pack 64 little-endian bytes into 8 words exactly.
- Proofs/ScalarBytesSpec.lean (3308 lines): bytes_word_loop_spec_0..7,
each split head/tail at j=4 (the 8-fold monolith grows exponentially
in elaboration - METHOD 4). Disjoint-bit ORs become additions via
core's Nat.two_pow_add_eq_or_of_lt with explicit calc bridges (the
default simp set literalizes 2^8 -> 256 and breaks pow-form rewrites;
simp only everywhere).
- Proofs/ScalarUnpackSpec.lean: bytes_unpack_spec composes the eight
inner lemmas through the outer loop (iterator start needs a term-level
equality rewrite per peel).
The from_bytes_wide main walk itself is proven at elaboration level
(fail-probe verified end to end) but its single-decl kernel certificate
replays >30min; it ships next as a phase-split (plan in the control
repo's method notes). check-scalar.sh: 12 proof files, 12 kernel audits,
all exactly [propext, Classical.choice, Quot.sound]. Button green.
Canonicity pass (the layer is now closed under its own preconditions):
- sub_val_spec post carries the exact value equation
(exists beta <= 1, scVal r + scVal b = scVal a + ell*beta, with the
underflow guard beta = 1 -> scVal a < scVal b)
- add/montgomery_reduce/mul/aggregate posts all carry scVal r < ell:
canonical inputs give canonical outputs everywhere. Needed because
from_bytes_wide (hash-to-scalar) feeds Montgomery outputs into add.
Hash-to-scalar foundation (toward Scalar::from_hash / EdDSA verify):
- extraction scope + from_bytes_wide (brings constants::R); regenerated gen
- source repos carry a documented Aeneas-compat patch: the bare
`hi[4] = words[7] >> 20` extracts ill-typed at pin bf13c42e; masked
(semantic no-op, words[7] >> 20 < 2^44)
- Proofs/ScalarWideSpec.lean: R constant lemmas (R = 2^260 mod ell,
witness 2^260 = R + 255*ell) and montgomery_mul_spec, the single
Montgomery round: [r]*2^260 = [a]*[b], canonical bounded output
check-scalar.sh: 10 proof files, 11 kernel audits, all exactly
[propext, Classical.choice, Quot.sound]. Button pressed fresh: green.
The solana fork's Scalar52 sub/add/conditional_add_l extract token-identical
to upstream dalek (only the crate namespace differs: curve25519 vs
curve25519_dalek), so ScalarSubSpec/ScalarAddSpec port with the namespace
adjustment and verify against THIS fork's own gen (R2). ScalarLoop
infrastructure included. check-scalar.sh at dalek parity: full manifest +
5/5 kernel axiom audit, green at 300s/4096MB.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- check.sh: proofs memory default 6144 -> 8192 (ReduceSpec's norm_num
step peaks above 6144; guard aborted gracefully — R3 was broken, S1
held). Matches pasta's calibration.
- check.sh: dead-file gate now exempts Scalar* (delegated to
check-scalar.sh); the gate had been un-passable since the scalar layer
landed, masked by the memory failure.
- check.sh: axiom-audit phase routed through lean-guard (cgroup + flock;
was raw lean -M), audit temp file moved into the workspace (lake env
rejects /tmp inputs — the /tmp phase had never run green).
- check-scalar.sh: NEW Phase 3 kernel axiom audit — ScalarProofs.L_val
must report exactly [propext, Classical.choice, Quot.sound].
- README: signature layer '⏳ planned' (was 'in progress' with nothing
started); planned certificate names marked as such.
Validated: full check.sh + check-scalar.sh green end-to-end in the pass-2
sweep (see formal-verification-control/COHERENCE-PASS-2.md).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Transpile the Scalar52 limb backend (backend::serial::u64::scalar
add/sub/mul/square/montgomery_*) from Rust to Lean via Charon/Aeneas,
scoped at the function level to the iterator-free arithmetic core.
- verification/extract-scalar.sh: function-level Charon/Aeneas extraction
- verification/gen/CurveScalar/{Types,Funs}.lean: transpiled model (28 defs)
- verification/gen/CurveScalar/{TypesExternal,FunsExternal}.lean: hand-written
external models (subtle.Choice + 2 subtle fns; namespace = curve25519)
- verification/Proofs/ScalarDenote.lean: semantic foundation — Scalar52
denotation into ℤ/ℓℤ, limb-bound invariant, and L_val (the transpiled
constants::L denotes exactly the group order ℓ, kernel-checked)
- verification/check-scalar.sh: guarded compile of the four gen modules
plus the denotation foundation
check-scalar.sh passes: gen compiles; denotation + L = ℓ proven.
add/sub/mul remain in progress.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Extraction widened to backend::serial::curve_models + edwards (matching the
reference recipe; extra opaque: backend::scalar_fits_in_128_bits — a
post-reference NAF-path helper whose generated code trips an Aeneas
namespace-shadowing wart). Reference Ed* suite compiles UNCHANGED (same
crate namespace). All proofs pass; both certificates axiom-clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
