Commit Graph
15 Commits
Author SHA1 Message Date
N0zoM1z0morlutofactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>Kaoru0822
14c93a753f fix(perf): validate summary coverage and normalized provenance
Reject PCAP summaries that do not describe their single complete view and
reject present normalized limitations/unknowns outside string arrays at the
shared observation boundary and factual grader. Preserve omission and empty
arrays. Add regressions that reproduce faithful malformed receipts and keep
original model measurements unchanged.

Validation: 28 context/grader tests, seven actual SDK consumer fault/control
probes, and factual regrading of all four original model answers.

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
2026-10-14 02:40:31 +08:00
N0zoM1z0morlutofactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>Kaoru0822
620ac0f5a7 fix(perf): reject malformed PCAP provenance before grading
Parse present summary/page containers independently of the claimed kind so
malformed nested values write failed facts rather than crashing. Validate
outer limitations as arrays of strings before producing observations and
admitting citations. Preserve empty valid lists and original measurements.

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
2026-10-14 02:40:31 +08:00
N0zoM1z0morlutofactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>Kaoru0822
0f24c61196 fix(perf): report empty PCAP observation lists as failed facts
Reject an empty observation list at citation admission before reading its
first receipt, and keep writing failed factual reports. Add the full-grader
regression reproducing the prior IndexError.

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
2026-10-14 02:40:31 +08:00
N0zoM1z0morlutofactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>Kaoru0822
8924a3a06c fix(perf): bind PCAP receipts to requested parameters and views
Preserve completed call arguments with receipt ownership in the grader, verify
actual call and envelope parameters, and validate delegated summary/URI views
once at the common observation boundary. Preserve existing delegation and
original measurements. Add complete-grader regressions for request mismatch.

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
2026-10-14 02:40:31 +08:00
N0zoM1z0morlutofactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>Kaoru0822
f07446681c test(perf): validate PCAP receipt IDs and exact view types
Reject malformed response identities at observation production and preserve
JSON scalar types when checking requested and cited PCAP views. Keep normal
complete and delegated summary workflows unchanged.

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
2026-10-14 02:40:31 +08:00
N0zoM1z0morlutofactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>Kaoru0822
6f0237af0c fix(experiments): reject malformed PCAP receipts as failed facts
Parse required receipt and nested view containers before indexing, preserving unsupported/conflicting identities as unknown instead of raising while grading. Compare cited page coverage and same-ID envelope conflicts as JSON to preserve boolean, integer and float distinctions. Reproduce prior crashes and false passes, then verify 22 context/grading and six comparator regressions plus factual regrading of all four original model answers without reruns. SDK and production contracts remain unchanged.

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
2026-10-14 02:40:31 +08:00
N0zoM1z0morlutofactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>Kaoru0822
3e88ebf735 fix(experiments): preserve PCAP receipt producer provenance
Keep tool and Evidence together in the receipt index, reject conflicting ID ownership, and validate requested tool and delegated operation roles. Preserve legitimate summary delegation to inspect_analysis_view and identical repeated receipts. Validate the same operation contract at the SDK observation boundary. Twenty grading/context tests and six comparator tests pass after reproducing the gap; real SDK fault/control runs and all four stored actual-model factual grades pass without model reruns.

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
2026-10-14 02:40:31 +08:00
N0zoM1z0morlutofactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>Kaoru0822
073f6c07e9 fix(experiments): validate PCAP receipt ownership before observation
Validate subjects and normalized targets against selected bytes/path at the common PCAP observation boundary. Require every retained view parent to be linked and distinct from the response identity; enforce the same relationship for model citations. Remove redundant page target checks. Actual SDK before/after faults and normal complete/summary controls pass, as do 18 grading/context and six comparator regressions plus factual regrading of all four stored model answers. No model rerun or production/default change.

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
2026-10-14 02:40:31 +08:00
N0zoM1z0morlutofactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>Kaoru0822
d688507563 fix(experiments): validate PCAP coverage and cited artifact subjects
Verify complete page coverage at the SDK measurement boundary. Bind every cited Evidence subject digest and local path to the reference capture in the model grader, and reuse the reference page coverage. Add failing-then-passing regressions, real SDK consumer fault/control checks, and factual regrading of all four stored model answers without model reruns. Original measurements and resource uncertainty stay intact.

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
2026-10-14 02:40:31 +08:00
N0zoM1z0morlutofactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>Kaoru0822
6b167f9812 fix(experiments): bind PCAP prefix coverage and requested view
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
2026-10-14 02:40:31 +08:00
N0zoM1z0morlutofactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>Kaoru0822
2a7982ce2b fix(experiments): verify cited PCAP metadata and exact receipt IDs
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
2026-10-14 02:40:31 +08:00
N0zoM1z0morlutofactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>Kaoru0822
3c36892ab7 fix(experiments): bind PCAP facts to the cited receipt chain
Validate URI facet/range/content and shared retained parent instead of accepting receipt kinds alone. Exercise the complete grader with deterministic positive and malformed/wrong-receipt cases; close transcript file handles. All captured real answers still pass without model reruns.

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
2026-10-14 02:40:31 +08:00
N0zoM1z0morlutofactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>Kaoru0822
fb10bdc546 test(perf): compare complete and retained-page PCAP workflows
Pin official public HTTP recordings and a fixed metadata/URI question. Preserve each delivery scope, accept equivalent explicit model labels without losing contradictions, and verify related grading regressions. This adds experiments only; production and complete defaults stay unchanged.

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
2026-10-14 02:40:31 +08:00
记得吃中饭andGitHub 9c14999efd feat(ghidra): support raw NMOS 6502 analysis (#2079)
* feat(ghidra): admit raw NMOS 6502 analysis

* test(ghidra): require verified POSIX verifier cleanup

* test(ghidra): verify 6502 low-memory raw imports

* docs(skill): document raw NMOS 6502 support

* fix(ghidra): preserve requested Windows raw source paths

* docs(ghidra): include raw 6502 in Windows installation scope

* test(ghidra): verify Windows raw junction rejection through CLI and MCP

* fix(validation): make fixtures portable and preserve explicit Codex home

* fix(ghidra): attest raw mapping state and address space

* fix(ghidra): align diagnostics and regression groups

* fix(ghidra): bind raw attestation to language address space

* chore(ghidra): avoid verifier insertion conflict

* docs(ghidra): align linked raw admission guides

* fix(ghidra): retain 6502 verifier workspace on failure

* test(ghidra): isolate Windows authority doctor fixture

* fix(binary): re-admit Windows raw paths on session reopen

* test(binary): require fresh provider execution after path change

* Constrain 6502 to raw lifecycle and snapshot targets

* Bypass saved reads for unadmitted Windows paths

* Guard direct CLI snapshot replay on Windows paths

* fix(snapshot): gate Windows Ghidra replay on native admission

* fix(snapshot): constrain nested 6502 subjects to raw images

* test: make dyld slice fixtures portable
2026-10-13 21:13:01 +08:00
YsopwnableGitHubfactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
ee87111009 test(perf): distinguish Codex tool context and token accounting (#2416)
Add optional owned Codex tool/usage capture and exact-name discovery experiments while preserving default ephemeral trials and all REA public APIs. Retain scoped managed and actual Android results, mixed outcomes, source identities and resource unknowns.

Six semantic Python tests, full checked pre-push, byte-identical own-file rebase verification, manual review and all 39 exact-head CI checks pass. Historical main fixture failures and their independent repair remain documented; measured source/runtime pins are unchanged.

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
2026-10-13 09:01:17 +08:00