* feat(native): inspect recorded crashes through pwntools and pwndbg
* fix(native): bind core note semantics to raw owner bytes
* fix(native): validate recorded core ABI and evidence scalars
* fix(native): require complete recorded note interpretations
* fix(native): bind physical core notes to original headers
Bring in #927's merge of main (#1037 at skill version 30):
- README_ja/ar/ko/zh: keep the merged tables and count
inspect_dyld_shared_cache in application workflows (13 -> 14).
- Skill: #927 now uses version 31, so this branch moves to 32.
- Regenerate the product catalog, conformance manifest, skill digest and
package metadata (138 tools).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
main now carries #1037 (skill version 30, reworded native binary row):
- README_ja/ar/ko/zh: take main's tables and count
trace_dylib_resolution in the artifact graph row (5 -> 6).
- Skill: move to version 31.
- Regenerate the product catalog, conformance manifest, skill digest and
package metadata (137 tools).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Bring in #927's merge of main (#964 at skill version 29):
- README_ja/ar/ko/zh: keep the merged tables and count
inspect_dyld_shared_cache in application workflows (13 -> 14).
- Skill: #927 now uses version 30, so this branch moves to 31.
- Regenerate the product catalog, conformance manifest, skill digest and
package metadata (137 tools).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
main now carries #964 (skill version 29, macOS native row 7 -> 8):
- README_ja/ar/ko/zh: take main's tables and count
trace_dylib_resolution in the artifact graph row (5 -> 6).
- Skill: move to version 30.
- Regenerate the product catalog, conformance manifest, skill digest and
package metadata (136 tools).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Resolve conflicts against main:
- package.json: ship both the LLDB tracer bridge and main's process run
token reader sources.
- README.md: take main's simplified English README, which no longer has
the tool-family table.
- README_ja.md: take main's reworded table and count
observe_native_calls in the macOS native row (7 -> 8).
- Skill: the references gain the observe_native_calls paragraph, so the
version moves to 29 (main already uses 28).
- Regenerate the product catalog, conformance manifest, skill digest and
package metadata (135 tools). The MCP tool catalog is now built rather
than committed (#997), so it is removed here.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Bring in #927's merge of main and resolve conflicts:
- README.md: main's simplified English README has no tool-family table.
- README_ja/ar/ko/zh: keep main's reworded tables and count
inspect_dyld_shared_cache in application workflows (13 -> 14).
- Skill: #927 now uses version 29, so this branch moves to 30.
- Regenerate the product catalog, conformance manifest, skill digest and
package metadata (136 tools). The MCP tool catalog is now built rather
than committed (#997), so it is removed here.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Resolve conflicts against main:
- README.md: take main's simplified English README, which no longer has
the tool-family table.
- README_ja.md: take main's reworded table and count
trace_dylib_resolution in the artifact graph row (5 -> 6).
- Skill: the references gain the trace_dylib_resolution paragraph, so
the version moves to 29 (main already uses 28).
- Regenerate the product catalog, conformance manifest, skill digest and
package metadata (135 tools). The MCP tool catalog is now built rather
than committed (#997), so it is removed here.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(mcp): advertise JSON values without recursive schemas
`z.json()` projects a self-referential `$defs/__schema0` entry. Twenty
tool input schemas and 123 output schemas carried it, and clients or
model APIs that reject recursive JSON Schemas refuse the whole tool
list. Every JSON value satisfies `{}`, so the shared JSON value schema
now advertises that. Runtime parsing still admits only JSON values.
Refs #919
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(mcp): keep advertised input schemas within model function-calling limits
Some model providers reject the whole tool list when any input schema
contains a recursive $ref or nests deeper than 10 levels (fixes#919).
Sanitize the advertised wire projection: break cyclic $refs, bound nesting
depth, and drop unreachable $defs. Canonical Zod contracts are untouched.
---------
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: sx4im <sx4im@users.noreply.github.com>
Co-authored-by: N0zoM1z0 <1705004398@qq.com>
`z.json()` projects a self-referential `$defs/__schema0` entry. Twenty
tool input schemas and 123 output schemas carried it, and clients or
model APIs that reject recursive JSON Schemas refuse the whole tool
list. Every JSON value satisfies `{}`, so the shared JSON value schema
now advertises that. Runtime parsing still admits only JSON values.
Refs #919
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: N0zoM1z0 <1705004398@qq.com>
* fix(browser): preserve authorized redirect hops
* fix(browser): compare redirect semantics without event times
* fix(browser): classify missing redirect predecessors
* fix(browser): distinguish malformed redirect URLs
* fix(browser): preserve evidence on malformed redirect envelopes
* fix(browser): freeze evidence after malformed redirect
* docs: regenerate catalogs after main integration
* fix(browser): account for skipped redirect body shapes
* fix(browser): mark skipped redirect response shapes unavailable
When a malformed redirect keeps a JSON predecessor, record that its response body shape was not captured.
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
* fix(contracts)!: require absolute local paths for remaining caller-supplied file inputs
Follow-up to the session snapshot/export tightening: the same invisible server-cwd resolution applied to open_binary target paths (parseBinaryTarget(path, process.cwd())), import_evidence_bundle bundle reads, firmware input files (resolve(raw)), browser executable launch paths (resolve-then-spawn), Electron executable/application/root paths (realpath on raw input), and the expected server path check.
Each field gains the shared isAbsoluteLocalPath refinement plus a catalog-visible description, since JSON Schema cannot express the rule. Examples, scenario fixtures, and contract tests updated to absolute paths; new schema cases cover rejection/acceptance/omission per field.
Explicitly unchanged (verified rooted or out of scope): inspect_plist paths (resolved against the target bundle dir), firmware output_directory (already provider-enforced), filesystem observation paths (resolved against the caller-supplied working_directory), keyed-archive member paths (archive-relative), JS evidence member paths (evidence-relative under root_alias), provider-internal execution schemas, CLI-only doctor flows, and CLI options (own resolve against the operator cwd).
* fix(cli): resolve operator-relative paths before shared absolute-path validation
The contract tightening rejects relative paths, but three CLI commands forward operator-relative input into the shared schemas: firmware commands pass raw args to FirmwareAnalysisService, and the Electron/browser scenario commands validate parsed scenario JSON directly. CLI paths are relative to the operator shell, so resolve them against the working directory first: resolve() in firmwareCommands, and a shared resolveCliJsonPaths JSON key-path helper (cliJsonInput) applied in electronCommands and cliBrowserScenarioCommands. MCP behavior is unchanged — MCP callers must still send absolute paths.
Covers the regression with a relative-path firmware CLI acceptance test (asserts provider-stage capability_unavailable, not a validation error) and unit tests for the helper.
fix(contracts): require absolute local paths for session filesystem inputs
Snapshot, evidence-bundle export, and managed-target inputs accepted any non-empty string, so relative paths were resolved against the MCP server working directory instead of the caller's intent. Refine these contract fields to absolute host paths (Windows drive and forward-slash forms included), advertise the constraint on the property descriptions, and regenerate the derived catalog, product catalog, and skill catalog digest. Relative CLI options are unaffected because CLI flows do not parse these MCP contracts.
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
observe-native-calls now runs through composition/directAnalysis, where #965
moved runProviderAnalysis.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Add observe_native_calls (CLI: observe-native-calls <path> <input-json>). It
launches the active Mach-O as an owned process under `lldb --batch`, stops at
caller-selected functions or Objective-C methods, records each entry and
auto-continues:
- thread, image, symbol, load and file address;
- raw integer argument registers and optional caller frames;
- for methods, the selector read from _cmd and the receiver's dynamic class,
from LLDB's runtime reader without running target code.
No expression is evaluated and no existing process is attached. The run ends
on exit, max_events or duration_ms; REA then kills the process and confirms
that it is gone. The result also carries exit status, bounded stdout/stderr,
breakpoint resolution with partial coverage for unresolved requests, and
limitations. A target LLDB may not debug fails with a debugger-attach-denied
reason and remediation.
verify:native-calls checks the tool through the CLI and stdio MCP against a
compiled fixture, compares breakpoint addresses with nm, exercises every
outcome, and checks hardened-runtime targets with and without get-task-allow.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Carry conditional-load propagation, parsed-image digest binding, and
segment-wise root resolution into the shared-cache trace. Also:
- Make cache compatibility directional: an arm64e cache serves arm64
processes and an x86_64h cache serves x86_64 ones, not the reverse.
- Require the complete 24-byte LC_BUILD_VERSION with its declared tool
records, and the 16-byte LC_VERSION_MIN_* command, before recording a
platform.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Edges below a conditional fallback now carry loader_conditional, reuse of
a conditional image by install name is conditional, and their findings say
they apply only if that image loads.
- Hash each traced image only if it is still the file whose headers were
parsed (device, inode, size and change times), before and after hashing.
- Resolve explicit roots segment by segment, so a root below a symlinked
directory that points outside the analyzed root is rejected.
- Require a complete 16-byte LC_CODE_SIGNATURE command.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Use a supplied cache for a process only when it serves the process's CPU
family and one of its LC_BUILD_VERSION or LC_VERSION_MIN platforms matches
the cache's platform or alternate platform; slices now report platforms.
A limitation names each root the cache was not used for, and why.
- Treat an image listed in the cache but mapped in a missing or
UUID-mismatched subcache as undetermined, list it in
coverage.unverified_shared_cache_images, and mark coverage partial.
- Keep a cache hit conditional when an earlier candidate is undeterminable.
- Open a development cache's named subcaches beside its base name.
- Describe absolute paths as not evaluated only when no cache is supplied.
- Preflight dyld_info in verify:dyld-shared-cache.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
inspect_native_dispatch_metadata read pointer fields as stored bytes, so
chained-fixup images (the default since Xcode 13) decoded no classes, and
superclasses from other images were always unresolved.
- Decode LC_DYLD_CHAINED_FIXUPS pointer formats 64, 64_OFFSET, ARM64E,
ARM64E_USERLAND and ARM64E_USERLAND24 with import formats 1-3, and
interpret LC_DYLD_INFO bind, weak-bind and lazy-bind opcode streams.
- Name external superclasses from their _OBJC_CLASS_$_ binds.
- Decode class_ro_t baseProperties and __objc_catlist category_t records
(instance/class methods, protocols, properties) into objc_categories;
category methods carry their category name.
- Report the fixup encoding in a pointer_fixups coverage facet.
- verify:apple-dispatch builds the fixture with dyld-info and chained
fixups, plus arm64e on Apple silicon.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Address review findings in trace_dylib_resolution:
- Resolve LC_LAZY_LOAD_DYLIB dependencies without traversing them, since
dyld loads them on first use. An unresolved one becomes
`lazy-load-unresolved`, not a launch failure.
- Accept FAT tables up to 128 records, as target opening does. A table
that names no Mach-O CPU type, such as a Java class file, is not
Mach-O.
- Require 8-byte load-command alignment in 64-bit images. 1,508 real
slices still match otool.
- Report EACCES/EPERM while probing, reading, or hashing as an
`unavailable` failure naming the code and path, rather than generic
io.
Findings and limitations move to dylibResolutionFindings.ts to keep the
resolver under the module size limit.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
macOS 11 and later ship system libraries only inside the dyld shared
cache. trace_dylib_resolution therefore left every /usr/lib and /System
load undetermined, and REA had no way to see what a cache contains.
inspect_dyld_shared_cache (CLI inspect-dyld-shared-cache) reads a cache
file set selected by path, without an open target. Its pure TypeScript
reader parses modern and legacy dyld_cache_header layouts and reports:
- UUID, architecture, platform, OS version, Mac Catalyst alternate,
cache type, shared region, max slide;
- every VM mapping;
- each subcache by file suffix, checked by UUID. A missing or
mismatched subcache makes coverage partial;
- the complete image list.
For each requested install path it reads that image's load commands
through the cache mappings with the existing Mach-O reader.
trace_dylib_resolution accepts an optional shared_cache. Absolute
candidates found in a cache that serves the process's CPU family resolve
with the new outcome and status `shared-cache`. That status also takes
precedence over later in-bundle fallbacks such as a back-deployed
libswiftCore. The result names the cache UUID and OS version it used.
verify:dyld-shared-cache checks every image in the host cache (3,649 on
macOS 26.6), with dependencies and their re-export, weak, upward and
delay-init attributes, against `dyld_info -all_dyld_cache
-linked_dylibs`. It also checks CLI/MCP parity, and that each
shared-cache edge in the fixture trace is a library dyld actually loads.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Address review findings in trace_dylib_resolution:
- Choose child slices by dyld's graded architectures, so an x86_64h
process loads a generic x86_64 slice instead of reporting
architecture-missing.
- Read FAT containers with the byte-swapped FAT_CIGAM and FAT_CIGAM_64
magics, which target opening already accepts.
- Report string-bearing load commands too short to hold their string
offset as malformed instead of failing with a RangeError.
- Keep recognized but unsupported images as an `unsupported` candidate
outcome. It is undeterminable like an outside-root path, not
malformed.
- List Mach-O files that do not parse during root enumeration as
unparsed images, so coverage is partial rather than silently complete.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Add trace_dylib_resolution (CLI trace-dylib-resolution) to the artifact
family. For the active Mach-O, or every executable in the active .app
(the main app, XPC services, extensions, login items, helpers), it
reports which file each dylib load reaches and how.
A pure TypeScript reader decodes FAT and Mach-O load commands:
LC_LOAD/WEAK/REEXPORT/LAZY/UPWARD_DYLIB, including the newer
dylib_use_command encoding with weak, re-export, upward and delayed-init
flags, plus LC_ID_DYLIB, LC_RPATH, LC_DYLD_ENVIRONMENT and code-signature
presence. Malformed tables are reported with the failing command instead
of being guessed.
Resolution follows dyld:
- @executable_path comes from each process root;
- @loader_path comes from the loading image, or from the owner of the
LC_RPATH entry being expanded;
- @rpath searches the loader's rpaths first, then up the load chain;
- dependencies load dependents-first in load-command order, once per
process, and a matching install name reuses an already loaded image;
- symlinks are followed inside the analyzed root, and escapes stop.
Every candidate path is reported with its outcome. Each edge resolves to
resolved, conditional (an outside-root candidate such as /usr/lib/swift
precedes), unresolved, or undetermined. Absolute and shared-cache system
paths are never checked against the host. Derived findings cover
unresolved required or weak loads, absent earlier @rpath candidates (with
library validation left unknown), and LC_DYLD_ENVIRONMENT.
verify:macos-bundle now also:
- checks every traced image against otool -l;
- asserts each resolution status on a clang-built fixture that adds a
missing Overrides rpath, an absolute rpath, and a delay-init library;
- compares the predicted load order of two process roots with what dyld
loads under DYLD_PRINT_LIBRARIES.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
project_apple_application_graph accepted only IPA inventories. It now
also projects macOS .app directories and ZIP or DMG inventories that
contain applications. It reports application roots, platforms, and
every nested bundle with its path-convention role, layout, parent,
Info.plist path, executable candidates, and signing paths:
- roles: app extensions, XPC services (including inside frameworks),
frameworks, login items, system and driver extensions, plug-ins,
resource bundles, and helper or nested applications;
- privileged helpers in Contents/Library/LaunchServices and launchd
plists in Contents/Library/LaunchAgents or LaunchDaemons;
- direct files of Contents/Helpers;
- symlinks by path. Their targets are not inventoried, so versioned
frameworks are described through their real Versions/<version>/
directories.
AppleDouble sidecars (._* and __MACOSX/) are excluded from roles. A DMG
whose contents could not be inventoried is reported as partial. Other
inventory formats remain input errors. The projection still reads no
plist or CMS semantics; it points callers at inspect_plist.
The result schema is additive. projection_id now covers the new fields,
so IPA projection IDs differ from earlier versions.
verify:macos-bundle compiles a source-owned app with clang (CLT only).
It packs it as a directory, a ditto ZIP, and an APFS DMG, and checks
identical anatomy through the CLI with stdio MCP parity. It runs in the
macOS CI job.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* perf(graph): avoid repeated traversal edge scans and copies
* fix(android)!: preserve JVM settings and reuse metadata sessions
Read class metadata without generating application source, retain immutable serialized APK sessions, and join owned cleanup on cancellation and shutdown.
BREAKING CHANGE: Android analysis now requires a full JDK 17 or newer with its compiler module for the packaged Java source bridge.
A field that changed from a decodable to an undecoded signature was
reported as removed on the decoded side and unknown on the other. The
undecoded member with the same declared type and name may be that field
with a changed signature, so its decoded counterpart is now unknown too,
with a counterpart-signature-not-decoded limitation.
The compare_managed_members description now names the exact declared
type, name, and raw signature tier and the undecoded handling.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(browser): trace captured website module imports
Compose verified exported scripts, inert native import syntax and a narrow
Chromium resolution port. Preserve query/fragment identity, import-map context,
all reported capture candidates and explicit unknown execution. Share named
CLI/MCP contracts and retain native reports, diagnostics and cleanup failures.
Add focused boundary regressions and a real module-loading oracle to the
existing conditional Chrome lane. No captured application code is executed
by module tracing, and no selected website asset is refetched.
Refs #851, #774
Assisted-by: Yukari (gensokyo-skills:yukari-boundary-analysis)
Assisted-by: Kogasa (gensokyo-skills:kogasa-surprise-testing)
* fix(browser): preserve module contexts and source bindings
Keep capture completeness inline, validate actual source text against its
digest, and separate unsupported URL contexts from malformed input.
Use the importer origin for native module resolution and the selected map
base independently, including mixed HTTP/HTTPS and escaped HTML attributes.
Refs #851, #774
Assisted-by: Yukari (gensokyo-skills:yukari-boundary-analysis)
Assisted-by: Kogasa (gensokyo-skills:kogasa-surprise-testing)
morluto
Ysopwnable
github-actions[bot]
Kaoru0822
Saim Shafique
Dery Ferdika