ltl-accumulator-verified/verification/inventory_gate.sh

114 lines
5.9 KiB
Bash
Raw Permalink Normal View History

Review round 3: environment-derived audit surface, self-contained kit Round-2 external reviews (GPT-5.6 + second Claude) converged on the coverage gate being evadable (H1/NEW-1); GPT additionally proved the kit's fidelity target could not run (H2) and the namespace-collision attack that defeats any source-regex fix. This round adopts GPT's required correction in full: - Proofs/Inventory.lean: declaration inventory read from the compiled Lean environment — every constant of every corpus module, fully qualified, unfiltered (compiler auxiliaries and _private mangles pinned too), with kind and axiom cone; own cone walker cross-checked in-process against core collectAxioms (hard error on divergence). - verification/inventory-allowlist.txt: all 218 constants pinned. - inventory_gate.sh: fail-closed diff both directions (UNCLASSIFIED / STALE), INV-COUNT truncation guard, exactly-one-axiom invariant. - check.sh Phase 3b rewritten around the gate + manifest⇔inventory drift checks + CONES⇔inventory cone cross-check (two independent computations must agree). EXCLUDE table gone (sha256/Bytes are ordinary audited entries now). - selftest_audit.sh: 9 adversarial cases against the production gate (attributed/indented/private/instance, namespace collision, smuggled axiom, deleted decl, unmanifested Proofs/ and gen/ modules) + positive control — all defeated (GPT release condition 2). - M1: recursive orphan-olean guard (caught a stray dev artifact on its first run), gen/ dead-file check, corpus-wide single-axiom pin. - L1/NEW-2: acceptIncl_sound drops the redundant hm (derived from hacc.1); cone unchanged. - M2/M3: STATEMENT-MAP counts 230,271/230,016; non-vacuity guard wording narrowed to what the guards actually certify. - README layer table: stale L4/pin-store rows fixed (missed by both round-2 reviewers AND the round-2 revision — found in self-review). - KNOWN-GAPS 12 (audit-gate lineage + residual limits), 13 (round-2 kit target not self-contained); gap 2 count fixed. - RESPONSE-TO-REVIEWERS.md: round-3 disposition of every finding. Kit round 3 additionally ships the complete stdlib-only import closure of pacta.transparency (content-addressed vs pacta 3d81d53), the clean-extraction fidelity transcript (exit 0, 230,271+230,016, zero mismatches), the ATTESTATION GREEN check.sh transcript, and the self-test transcript. The live LTL remains untouched (12 leaves, root bcd15f9d…); attestation stays blocked pending ePrint decision + author review + explicit operator order. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-11 22:32:18 +00:00
#!/usr/bin/env bash
# ─────────────────────────────────────────────────────────────────────────────
# inventory_gate.sh — diff an observed environment inventory against the
# pinned allowlist. This is THE production coverage gate: check.sh Phase 3b
# calls it, and selftest_audit.sh exercises this exact script against
# injected evader declarations — the tested logic IS the shipping logic.
#
audit: bind the scalar statements, and make the accounting identity mean audit 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>
2026-08-03 10:15:26 +00:00
# Usage: inventory_gate.sh <observed-lean-output> <allowlist-file> [<tag>]
#
# <tag> defaults to INV — the CORPUS walk. Pass DRV to gate the INSTRUMENTS'
# OWN SURFACE with this same implementation.
#
# WHY THE TAG EXISTS — round-8 review (Claude, register keys
# `drv-surface-no-cones`, `accounting-certifies-enumeration`).
#
# The accounting identity added in round 7 proved every constant the kernel
# sees is ENUMERATED by one of the two walks. The reviewer showed that
# enumeration is not audit: a claim planted in an instrument WAS enumerated —
# `DRV|LTLAccAudit.bait.smuggled|theorem` — and then nothing looked at it,
# because DRV rows carried name and kind and NO CONE, and no allowlist covered
# them. In their words, the identity "converted 36 declarations nobody
# enumerated into 36 declarations nobody examined. That is progress of one
# step, not two."
#
# The second step is here: DRV rows now carry their axiom cone and are pinned
# in a committed allowlist, by THIS gate, in both directions — exactly as the
# corpus is. One implementation, not two, because a second copy of a coverage
# gate is a second thing to drift.
#
# It also retires a heuristic. The driver-surface rule permits a theorem whose
# name extends a constant declared alongside it, since that is what the
# elaborator generates for a definition; the reviewer showed it "breaks in one
# line" — declare `def bait`, then `theorem bait.smuggled` passes. That rule is
# kept as a fast, readable first line of defence, but it is NO LONGER
# LOAD-BEARING: a planted claim now has to appear in the pinned allowlist, and
# a new row fails closed whatever it is named.
Review round 3: environment-derived audit surface, self-contained kit Round-2 external reviews (GPT-5.6 + second Claude) converged on the coverage gate being evadable (H1/NEW-1); GPT additionally proved the kit's fidelity target could not run (H2) and the namespace-collision attack that defeats any source-regex fix. This round adopts GPT's required correction in full: - Proofs/Inventory.lean: declaration inventory read from the compiled Lean environment — every constant of every corpus module, fully qualified, unfiltered (compiler auxiliaries and _private mangles pinned too), with kind and axiom cone; own cone walker cross-checked in-process against core collectAxioms (hard error on divergence). - verification/inventory-allowlist.txt: all 218 constants pinned. - inventory_gate.sh: fail-closed diff both directions (UNCLASSIFIED / STALE), INV-COUNT truncation guard, exactly-one-axiom invariant. - check.sh Phase 3b rewritten around the gate + manifest⇔inventory drift checks + CONES⇔inventory cone cross-check (two independent computations must agree). EXCLUDE table gone (sha256/Bytes are ordinary audited entries now). - selftest_audit.sh: 9 adversarial cases against the production gate (attributed/indented/private/instance, namespace collision, smuggled axiom, deleted decl, unmanifested Proofs/ and gen/ modules) + positive control — all defeated (GPT release condition 2). - M1: recursive orphan-olean guard (caught a stray dev artifact on its first run), gen/ dead-file check, corpus-wide single-axiom pin. - L1/NEW-2: acceptIncl_sound drops the redundant hm (derived from hacc.1); cone unchanged. - M2/M3: STATEMENT-MAP counts 230,271/230,016; non-vacuity guard wording narrowed to what the guards actually certify. - README layer table: stale L4/pin-store rows fixed (missed by both round-2 reviewers AND the round-2 revision — found in self-review). - KNOWN-GAPS 12 (audit-gate lineage + residual limits), 13 (round-2 kit target not self-contained); gap 2 count fixed. - RESPONSE-TO-REVIEWERS.md: round-3 disposition of every finding. Kit round 3 additionally ships the complete stdlib-only import closure of pacta.transparency (content-addressed vs pacta 3d81d53), the clean-extraction fidelity transcript (exit 0, 230,271+230,016, zero mismatches), the ATTESTATION GREEN check.sh transcript, and the self-test transcript. The live LTL remains untouched (12 leaves, root bcd15f9d…); attestation stays blocked pending ePrint decision + author review + explicit operator order. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-11 22:32:18 +00:00
#
# Fail-closed in BOTH directions:
# UNCLASSIFIED — constant in the environment, absent from the allowlist
# (new/renamed decl, changed kind, or changed axiom cone)
# STALE — allowlist entry absent from the environment
# plus an output-integrity check: the INV-COUNT trailer emitted by
# Proofs/Inventory.lean must equal the number of INV lines actually seen,
# so a truncated or crashed run can never pass as an empty diff.
# ─────────────────────────────────────────────────────────────────────────────
set -uo pipefail
export LC_ALL=C # byte-order collation: sort/comm must agree with Lean's String order
audit: bind the scalar statements, and make the accounting identity mean audit 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>
2026-08-03 10:15:26 +00:00
obs_file="$1"; allow_file="$2"; TAG="${3:-INV}"
case "$TAG" in
INV) WHAT="the audited corpus"; TRAILER_TAG="INV-COUNT"; LABEL="inventory gate"; TRUNCLABEL="INVENTORY TRUNCATED" ;;
DRV) WHAT="the audit instruments"; TRAILER_TAG="DRV-COUNT"; LABEL="driver-surface gate"; TRUNCLABEL="DRIVER SURFACE TRUNCATED" ;;
*) echo " GATE MISUSE: unknown tag '$TAG' (expected INV or DRV)"; exit 1 ;;
esac
Review round 3: environment-derived audit surface, self-contained kit Round-2 external reviews (GPT-5.6 + second Claude) converged on the coverage gate being evadable (H1/NEW-1); GPT additionally proved the kit's fidelity target could not run (H2) and the namespace-collision attack that defeats any source-regex fix. This round adopts GPT's required correction in full: - Proofs/Inventory.lean: declaration inventory read from the compiled Lean environment — every constant of every corpus module, fully qualified, unfiltered (compiler auxiliaries and _private mangles pinned too), with kind and axiom cone; own cone walker cross-checked in-process against core collectAxioms (hard error on divergence). - verification/inventory-allowlist.txt: all 218 constants pinned. - inventory_gate.sh: fail-closed diff both directions (UNCLASSIFIED / STALE), INV-COUNT truncation guard, exactly-one-axiom invariant. - check.sh Phase 3b rewritten around the gate + manifest⇔inventory drift checks + CONES⇔inventory cone cross-check (two independent computations must agree). EXCLUDE table gone (sha256/Bytes are ordinary audited entries now). - selftest_audit.sh: 9 adversarial cases against the production gate (attributed/indented/private/instance, namespace collision, smuggled axiom, deleted decl, unmanifested Proofs/ and gen/ modules) + positive control — all defeated (GPT release condition 2). - M1: recursive orphan-olean guard (caught a stray dev artifact on its first run), gen/ dead-file check, corpus-wide single-axiom pin. - L1/NEW-2: acceptIncl_sound drops the redundant hm (derived from hacc.1); cone unchanged. - M2/M3: STATEMENT-MAP counts 230,271/230,016; non-vacuity guard wording narrowed to what the guards actually certify. - README layer table: stale L4/pin-store rows fixed (missed by both round-2 reviewers AND the round-2 revision — found in self-review). - KNOWN-GAPS 12 (audit-gate lineage + residual limits), 13 (round-2 kit target not self-contained); gap 2 count fixed. - RESPONSE-TO-REVIEWERS.md: round-3 disposition of every finding. Kit round 3 additionally ships the complete stdlib-only import closure of pacta.transparency (content-addressed vs pacta 3d81d53), the clean-extraction fidelity transcript (exit 0, 230,271+230,016, zero mismatches), the ATTESTATION GREEN check.sh transcript, and the self-test transcript. The live LTL remains untouched (12 leaves, root bcd15f9d…); attestation stays blocked pending ePrint decision + author review + explicit operator order. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-11 22:32:18 +00:00
audit: bind the scalar statements, and make the accounting identity mean audit 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>
2026-08-03 10:15:26 +00:00
# The trailer is an OUTPUT-INTEGRITY check: it must equal the number of rows
# the driver(s) actually emitted, BEFORE de-duplication. Comparing it to the
# de-duplicated count conflates "a run was truncated" with "two rows were
# identical", and the second is a record-format defect that must be fixed at
# the source, not absorbed here. (It was: DRV rows now carry their driver.)
N_RAW=$(grep -c "^$TAG|" "$obs_file" || true)
OBS=$(grep "^$TAG|" "$obs_file" | sort -u)
N_OBS=$(printf '%s' "$OBS" | grep -c "^$TAG|" || true)
if [ "$N_RAW" -ne "$N_OBS" ]; then
echo " DUPLICATE $TAG RECORDS: $N_RAW rows collapse to $N_OBS distinct ones."
echo " Two declarations share a record, so one is covered by the other's entry:"
grep "^$TAG|" "$obs_file" | sort | uniq -d | head -5 | sed 's/^/ /'
exit 1
fi
TRAILER=$(grep "^$TRAILER_TAG|" "$obs_file" | tail -1 | cut -d'|' -f2)
if [ -z "$TRAILER" ] || [ "$TRAILER" != "$N_RAW" ]; then
Review round 3: environment-derived audit surface, self-contained kit Round-2 external reviews (GPT-5.6 + second Claude) converged on the coverage gate being evadable (H1/NEW-1); GPT additionally proved the kit's fidelity target could not run (H2) and the namespace-collision attack that defeats any source-regex fix. This round adopts GPT's required correction in full: - Proofs/Inventory.lean: declaration inventory read from the compiled Lean environment — every constant of every corpus module, fully qualified, unfiltered (compiler auxiliaries and _private mangles pinned too), with kind and axiom cone; own cone walker cross-checked in-process against core collectAxioms (hard error on divergence). - verification/inventory-allowlist.txt: all 218 constants pinned. - inventory_gate.sh: fail-closed diff both directions (UNCLASSIFIED / STALE), INV-COUNT truncation guard, exactly-one-axiom invariant. - check.sh Phase 3b rewritten around the gate + manifest⇔inventory drift checks + CONES⇔inventory cone cross-check (two independent computations must agree). EXCLUDE table gone (sha256/Bytes are ordinary audited entries now). - selftest_audit.sh: 9 adversarial cases against the production gate (attributed/indented/private/instance, namespace collision, smuggled axiom, deleted decl, unmanifested Proofs/ and gen/ modules) + positive control — all defeated (GPT release condition 2). - M1: recursive orphan-olean guard (caught a stray dev artifact on its first run), gen/ dead-file check, corpus-wide single-axiom pin. - L1/NEW-2: acceptIncl_sound drops the redundant hm (derived from hacc.1); cone unchanged. - M2/M3: STATEMENT-MAP counts 230,271/230,016; non-vacuity guard wording narrowed to what the guards actually certify. - README layer table: stale L4/pin-store rows fixed (missed by both round-2 reviewers AND the round-2 revision — found in self-review). - KNOWN-GAPS 12 (audit-gate lineage + residual limits), 13 (round-2 kit target not self-contained); gap 2 count fixed. - RESPONSE-TO-REVIEWERS.md: round-3 disposition of every finding. Kit round 3 additionally ships the complete stdlib-only import closure of pacta.transparency (content-addressed vs pacta 3d81d53), the clean-extraction fidelity transcript (exit 0, 230,271+230,016, zero mismatches), the ATTESTATION GREEN check.sh transcript, and the self-test transcript. The live LTL remains untouched (12 leaves, root bcd15f9d…); attestation stays blocked pending ePrint decision + author review + explicit operator order. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-11 22:32:18 +00:00
echo " INVENTORY TRUNCATED: trailer=${TRAILER:-absent}, observed $N_OBS lines"
exit 1
fi
audit: bind the scalar statements, and make the accounting identity mean audit 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>
2026-08-03 10:15:26 +00:00
ALLOW=$(grep "^$TAG|" "$allow_file" | sort -u)
Review round 3: environment-derived audit surface, self-contained kit Round-2 external reviews (GPT-5.6 + second Claude) converged on the coverage gate being evadable (H1/NEW-1); GPT additionally proved the kit's fidelity target could not run (H2) and the namespace-collision attack that defeats any source-regex fix. This round adopts GPT's required correction in full: - Proofs/Inventory.lean: declaration inventory read from the compiled Lean environment — every constant of every corpus module, fully qualified, unfiltered (compiler auxiliaries and _private mangles pinned too), with kind and axiom cone; own cone walker cross-checked in-process against core collectAxioms (hard error on divergence). - verification/inventory-allowlist.txt: all 218 constants pinned. - inventory_gate.sh: fail-closed diff both directions (UNCLASSIFIED / STALE), INV-COUNT truncation guard, exactly-one-axiom invariant. - check.sh Phase 3b rewritten around the gate + manifest⇔inventory drift checks + CONES⇔inventory cone cross-check (two independent computations must agree). EXCLUDE table gone (sha256/Bytes are ordinary audited entries now). - selftest_audit.sh: 9 adversarial cases against the production gate (attributed/indented/private/instance, namespace collision, smuggled axiom, deleted decl, unmanifested Proofs/ and gen/ modules) + positive control — all defeated (GPT release condition 2). - M1: recursive orphan-olean guard (caught a stray dev artifact on its first run), gen/ dead-file check, corpus-wide single-axiom pin. - L1/NEW-2: acceptIncl_sound drops the redundant hm (derived from hacc.1); cone unchanged. - M2/M3: STATEMENT-MAP counts 230,271/230,016; non-vacuity guard wording narrowed to what the guards actually certify. - README layer table: stale L4/pin-store rows fixed (missed by both round-2 reviewers AND the round-2 revision — found in self-review). - KNOWN-GAPS 12 (audit-gate lineage + residual limits), 13 (round-2 kit target not self-contained); gap 2 count fixed. - RESPONSE-TO-REVIEWERS.md: round-3 disposition of every finding. Kit round 3 additionally ships the complete stdlib-only import closure of pacta.transparency (content-addressed vs pacta 3d81d53), the clean-extraction fidelity transcript (exit 0, 230,271+230,016, zero mismatches), the ATTESTATION GREEN check.sh transcript, and the self-test transcript. The live LTL remains untouched (12 leaves, root bcd15f9d…); attestation stays blocked pending ePrint decision + author review + explicit operator order. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-11 22:32:18 +00:00
FAILGATE=0
UNCLASS=$(comm -23 <(printf '%s\n' "$OBS") <(printf '%s\n' "$ALLOW"))
STALE=$(comm -13 <(printf '%s\n' "$OBS") <(printf '%s\n' "$ALLOW"))
if [ -n "$UNCLASS" ]; then
printf '%s\n' "$UNCLASS" | sed 's/^/ UNCLASSIFIED (in environment, not allowlisted): /'
FAILGATE=1
fi
if [ -n "$STALE" ]; then
printf '%s\n' "$STALE" | sed 's/^/ STALE (allowlisted, not in environment): /'
FAILGATE=1
fi
audit: bind the scalar statements, and make the accounting identity mean audit 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>
2026-08-03 10:15:26 +00:00
# THE AXIOM POLICY IS PER-SURFACE, and conflating the two would be a silent
# inversion. The CORPUS admits exactly one axiom — the sanctioned hash boundary.
# The INSTRUMENTS admit NONE: an audit tool that declares an axiom widens the
# trusted base from inside the thing doing the auditing. Reusing the corpus rule
# for DRV would have demanded LTLAcc.sha256 appear among instrument
# declarations, where it never does, and the gate would have failed for a reason
# that has nothing to do with the property.
Review round 3: environment-derived audit surface, self-contained kit Round-2 external reviews (GPT-5.6 + second Claude) converged on the coverage gate being evadable (H1/NEW-1); GPT additionally proved the kit's fidelity target could not run (H2) and the namespace-collision attack that defeats any source-regex fix. This round adopts GPT's required correction in full: - Proofs/Inventory.lean: declaration inventory read from the compiled Lean environment — every constant of every corpus module, fully qualified, unfiltered (compiler auxiliaries and _private mangles pinned too), with kind and axiom cone; own cone walker cross-checked in-process against core collectAxioms (hard error on divergence). - verification/inventory-allowlist.txt: all 218 constants pinned. - inventory_gate.sh: fail-closed diff both directions (UNCLASSIFIED / STALE), INV-COUNT truncation guard, exactly-one-axiom invariant. - check.sh Phase 3b rewritten around the gate + manifest⇔inventory drift checks + CONES⇔inventory cone cross-check (two independent computations must agree). EXCLUDE table gone (sha256/Bytes are ordinary audited entries now). - selftest_audit.sh: 9 adversarial cases against the production gate (attributed/indented/private/instance, namespace collision, smuggled axiom, deleted decl, unmanifested Proofs/ and gen/ modules) + positive control — all defeated (GPT release condition 2). - M1: recursive orphan-olean guard (caught a stray dev artifact on its first run), gen/ dead-file check, corpus-wide single-axiom pin. - L1/NEW-2: acceptIncl_sound drops the redundant hm (derived from hacc.1); cone unchanged. - M2/M3: STATEMENT-MAP counts 230,271/230,016; non-vacuity guard wording narrowed to what the guards actually certify. - README layer table: stale L4/pin-store rows fixed (missed by both round-2 reviewers AND the round-2 revision — found in self-review). - KNOWN-GAPS 12 (audit-gate lineage + residual limits), 13 (round-2 kit target not self-contained); gap 2 count fixed. - RESPONSE-TO-REVIEWERS.md: round-3 disposition of every finding. Kit round 3 additionally ships the complete stdlib-only import closure of pacta.transparency (content-addressed vs pacta 3d81d53), the clean-extraction fidelity transcript (exit 0, 230,271+230,016, zero mismatches), the ATTESTATION GREEN check.sh transcript, and the self-test transcript. The live LTL remains untouched (12 leaves, root bcd15f9d…); attestation stays blocked pending ePrint decision + author review + explicit operator order. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-11 22:32:18 +00:00
AXLINES=$(printf '%s\n' "$OBS" | grep '|axiom|' || true)
audit: bind the scalar statements, and make the accounting identity mean audit 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>
2026-08-03 10:15:26 +00:00
if [ "$TAG" = DRV ]; then
if [ -n "$AXLINES" ]; then
echo " AXIOM SURFACE DRIFT: $WHAT must declare no axioms; observed:"
printf '%s\n' "$AXLINES" | sed 's/^/ /'
FAILGATE=1
fi
AXNOTE="zero axioms declared by the instruments"
else
if [ "$AXLINES" != "INV|LTLAcc.sha256|axiom|LTLAcc.sha256" ]; then
echo " AXIOM SURFACE DRIFT: expected exactly LTLAcc.sha256, observed:"
printf '%s\n' "${AXLINES:- (none)}" | sed 's/^/ /'
FAILGATE=1
fi
AXNOTE="single sanctioned axiom"
Review round 3: environment-derived audit surface, self-contained kit Round-2 external reviews (GPT-5.6 + second Claude) converged on the coverage gate being evadable (H1/NEW-1); GPT additionally proved the kit's fidelity target could not run (H2) and the namespace-collision attack that defeats any source-regex fix. This round adopts GPT's required correction in full: - Proofs/Inventory.lean: declaration inventory read from the compiled Lean environment — every constant of every corpus module, fully qualified, unfiltered (compiler auxiliaries and _private mangles pinned too), with kind and axiom cone; own cone walker cross-checked in-process against core collectAxioms (hard error on divergence). - verification/inventory-allowlist.txt: all 218 constants pinned. - inventory_gate.sh: fail-closed diff both directions (UNCLASSIFIED / STALE), INV-COUNT truncation guard, exactly-one-axiom invariant. - check.sh Phase 3b rewritten around the gate + manifest⇔inventory drift checks + CONES⇔inventory cone cross-check (two independent computations must agree). EXCLUDE table gone (sha256/Bytes are ordinary audited entries now). - selftest_audit.sh: 9 adversarial cases against the production gate (attributed/indented/private/instance, namespace collision, smuggled axiom, deleted decl, unmanifested Proofs/ and gen/ modules) + positive control — all defeated (GPT release condition 2). - M1: recursive orphan-olean guard (caught a stray dev artifact on its first run), gen/ dead-file check, corpus-wide single-axiom pin. - L1/NEW-2: acceptIncl_sound drops the redundant hm (derived from hacc.1); cone unchanged. - M2/M3: STATEMENT-MAP counts 230,271/230,016; non-vacuity guard wording narrowed to what the guards actually certify. - README layer table: stale L4/pin-store rows fixed (missed by both round-2 reviewers AND the round-2 revision — found in self-review). - KNOWN-GAPS 12 (audit-gate lineage + residual limits), 13 (round-2 kit target not self-contained); gap 2 count fixed. - RESPONSE-TO-REVIEWERS.md: round-3 disposition of every finding. Kit round 3 additionally ships the complete stdlib-only import closure of pacta.transparency (content-addressed vs pacta 3d81d53), the clean-extraction fidelity transcript (exit 0, 230,271+230,016, zero mismatches), the ATTESTATION GREEN check.sh transcript, and the self-test transcript. The live LTL remains untouched (12 leaves, root bcd15f9d…); attestation stays blocked pending ePrint decision + author review + explicit operator order. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-11 22:32:18 +00:00
fi
audit: bind the scalar statements, and make the accounting identity mean audit 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>
2026-08-03 10:15:26 +00:00
[ "$FAILGATE" = 0 ] && echo " $LABEL: $N_OBS constants, environment == allowlist, $AXNOTE"
Review round 3: environment-derived audit surface, self-contained kit Round-2 external reviews (GPT-5.6 + second Claude) converged on the coverage gate being evadable (H1/NEW-1); GPT additionally proved the kit's fidelity target could not run (H2) and the namespace-collision attack that defeats any source-regex fix. This round adopts GPT's required correction in full: - Proofs/Inventory.lean: declaration inventory read from the compiled Lean environment — every constant of every corpus module, fully qualified, unfiltered (compiler auxiliaries and _private mangles pinned too), with kind and axiom cone; own cone walker cross-checked in-process against core collectAxioms (hard error on divergence). - verification/inventory-allowlist.txt: all 218 constants pinned. - inventory_gate.sh: fail-closed diff both directions (UNCLASSIFIED / STALE), INV-COUNT truncation guard, exactly-one-axiom invariant. - check.sh Phase 3b rewritten around the gate + manifest⇔inventory drift checks + CONES⇔inventory cone cross-check (two independent computations must agree). EXCLUDE table gone (sha256/Bytes are ordinary audited entries now). - selftest_audit.sh: 9 adversarial cases against the production gate (attributed/indented/private/instance, namespace collision, smuggled axiom, deleted decl, unmanifested Proofs/ and gen/ modules) + positive control — all defeated (GPT release condition 2). - M1: recursive orphan-olean guard (caught a stray dev artifact on its first run), gen/ dead-file check, corpus-wide single-axiom pin. - L1/NEW-2: acceptIncl_sound drops the redundant hm (derived from hacc.1); cone unchanged. - M2/M3: STATEMENT-MAP counts 230,271/230,016; non-vacuity guard wording narrowed to what the guards actually certify. - README layer table: stale L4/pin-store rows fixed (missed by both round-2 reviewers AND the round-2 revision — found in self-review). - KNOWN-GAPS 12 (audit-gate lineage + residual limits), 13 (round-2 kit target not self-contained); gap 2 count fixed. - RESPONSE-TO-REVIEWERS.md: round-3 disposition of every finding. Kit round 3 additionally ships the complete stdlib-only import closure of pacta.transparency (content-addressed vs pacta 3d81d53), the clean-extraction fidelity transcript (exit 0, 230,271+230,016, zero mismatches), the ATTESTATION GREEN check.sh transcript, and the self-test transcript. The live LTL remains untouched (12 leaves, root bcd15f9d…); attestation stays blocked pending ePrint decision + author review + explicit operator order. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-11 22:32:18 +00:00
exit "$FAILGATE"