This reverts commit 7336a566ac (#7119).
After #7120 and the other merges of 2026-10-25 grew inspect_analysis_view's
complete input schema (71,762 bytes plus 3,003 bytes of examples), the 64 KiB
fitting from #7119 could no longer keep instance level 3 and fell back to
level 2. That summarized the inline Evidence source and the view selector
collections, so main CI failed in toolInputReferences, toolSchemaValidity,
analysisViewMcp, absoluteLocalPathSchema, javascriptFactsViewMcp and
godotPckMcp. #7119's own CI ran against bc143125e, before that growth.
Reverting restores the complete full-profile advertisement while the size
limit is redesigned against these contracts (#4081 stays open).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Refs: #7135
* fix: abort invalid JSON parses and keep retained worker owners
Main (7375ab6e3) fails four boundary tests from its latest batch.
- f79fd59e9 moved JSON value validation into a non-aborting superRefine, so
authenticated Evidence still computed its recursive identity digest over
a 10,000-level `parameters` value and overflowed the stack. The process
comparison CLI then reported "Maximum call stack size exceeded" instead of
the depth limit. Abort the parse on a JSON validation issue so enclosing
refinements never walk invalid data.
- runOwnedCommand retains an incomplete cleanup through retainCleanup and
then rethrows the command failure without a cleanupOwner. The EVM and SQLite
providers replaced their retained owner with that undefined value, so an
output-limit failure with unresolved cleanup lost cleanupIncomplete and its
owner. Keep the retained owner unless the failure carries one.
- The EVM malformed-reply test expected a nonexistent `output_failure` code;
AnalysisOutputError projects to `unreadable_output`. The pcap capinfos stub
prints its report with a trailing newline, which retained output preserves.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(sqlite): keep a retained worker owner after a failed worker exit
Cover the path the existing worker-cleanup test bypasses: a real owned worker
exits nonzero, runOwnedCommand retains its incomplete cleanup and rethrows the
failure without a cleanupOwner. The provider must still report
cleanupIncomplete and block a new worker until the retained cleanup succeeds.
Fails on the previous owner assignment.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(evm): retain the compact worker failure beside retained cleanup
When the worker command fails and its cleanup is retained, the cleanup error
now reports the primary failure in the same compact form as an unretained
command cleanup failure: failure kind, message, exit code and signal. Its
output is already retained once as captured_output, so the previous error no
longer repeats up to a megabyte of diagnostics, and failure_kind stays at the
top level that the provider test reads. Both paths share
ownedCommandPreviousError.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* perf(artifacts): overlap bundle magic reads while enumerating dylib roots
trace_dylib_resolution opened every file of an app bundle one at a time
to read its Mach-O magic (Keynote: 35,971 files, 23 Mach-O images).
Keep up to eight reads in flight while the directory walk continues,
settling them in walk order so classification and failure precedence
match the sequential scan. Keynote 4.20 s -> 2.78 s.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(artifacts): pin ordered-read failure precedence and cleanup
Extract the overlapped magic-read queue into consumeWithOrderedReads and
test its ordering, concurrency bound, failure precedence, enumeration
close and read settlement with controlled promises.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* perf(native): skip llvm-nm's dyld-info pass when the symbol table names every export
inspect_macho ran `nm -gjU`, whose default dyld-info pass takes seconds on
large apps (Keynote arm64 5.8 s, x86_64 10.2 s) yet under -gU can only add
export-trie names that dyld_info -exports already reports. Run
`nm -gjU --no-dyldinfo` and rerun the default pass only when dyld_info
reports an export the symbol table lacks (e.g. strip -T), so its exact
names still disambiguate dyld_info rows.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(native): fall back to nm's dyld-info pass for unreadable export rows
Account for every dyld_info export line instead of only the rows the
parser returns, so an unfamiliar row format reruns the default nm pass.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* perf(managed): add retained member summary and pages
inspect_managed_members always returned the assembly-wide member_refs, so
one selected or unmatched type produced megabytes, and an unselected
ordinary assembly exceeded the MCP response budget. Add opt-in
detail: "summary" that retains the complete Evidence and returns identity,
selection with unmatched names, coverage, user-string coverage, limitations
and per-collection counts. Page original rows through inspect_analysis_view
with kind managed-members; each page names its collection so no row schema
is chosen by union order. Complete delivery stays the default.
Refs #3010
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* perf(managed): validate a sealed member parent once across pages
Each managed-members page reparsed the whole parent report, about 35 ms
for an 18 MB assembly result regardless of the page size. The analysis
view service now reports whether the parent Evidence is a sealed
snapshot, and the member view reuses the validated report for that
sealed result through a weak reference. Unsealed input is still
validated on every call.
The round-trip test fixture now carries an exported type, an ldstr
literal, a field access and a call edge, and requires every collection
to be nonempty so a row schema that drops facts cannot pass on empty
arrays.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(managed): keep delivery detail out of retained Evidence identity
Summary, omitted and complete delivery of the same managed inspection
now retain one parent record, as inspect_macho already does.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(native): derive Mach-O word size and byte order from the selected slice
inspect_macho read endian and word_size from `file -b` for the whole
file. macOS `file` never prints a Mach-O byte order, so endian was null
for every target. For a universal binary, word_size matched "64-bit"
anywhere in the summary instead of describing the selected slice, and it
was null for a thin 32-bit slice.
Both now come from the selected slice's otool header: the magic gives
the word size and the CPU family gives the byte order (x86 and ARM
little, PowerPC big, other families unknown). The file capture is no
longer needed by inspect_macho, so it is no longer run there.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(native): select a 32-bit slice beside a 64-bit slice through inspect_macho
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(native): classify every Swift mangling prefix as invalid when undecodable
demangle_swift reported "invalid" only for $s and _$s inputs that
swift-demangle left unchanged. Undecodable symbols carrying the other
prefixes the Swift demangler accepts (_T0, $S, _$S, @__swiftmacro_, $e,
_$e) were reported as "unchanged", i.e. as non-Swift names. Legacy _T
stays "unchanged" because it is also an ordinary C symbol spelling.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(native): expect only toolchain-portable Swift manglings to demangle
The macOS 15 runner's swift-demangle predates Embedded Swift, so valid
$e symbols stay unchanged there. Keep $e and _$e in the undecodable
half only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(capture): end special-scheme URL authorities at a backslash
The native mitmproxy adapter removed transport userinfo with
`^(...//)[^/?#]*@`, so the authority ran to the last "@" before "/", "?"
or "#". For `https://example.test\part@tail` it treated
`example.test\part` as userinfo, reported `https://tail` as the
Location, and withdrew the original bytes as a transport credential.
WHATWG URL parsing, used by the shared TypeScript owner
(recordedTransportUrl.ts) and by HAR captures, ends a special-scheme or
scheme-relative authority at a backslash, so that "@" is ordinary path
text.
Locate the authority with the same boundary rule before removing
userinfo. Bytes are handled as Latin-1 text so unrelated bytes stay
identical. Userinfo before the backslash is still removed and redacted.
Add source unit tests for the adapter projection to test:python.
Fixes#2671
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(capture): skip special-scheme leading slashes before userinfo
Ending the authority at a backslash exposed a credential: for
`https://\user:pass@host/path`, WHATWG parsing skips every leading "/"
or "\" after a special scheme (and in scheme-relative references), so
`user:pass` is userinfo, but the new boundary stopped at the first
backslash and found no "@". The same applied to `https:\\user:pass@host`
and `https:user:pass@host`.
Start the authority after that whole slash run for non-file special
schemes and scheme-relative references. File URLs keep their "//"
boundary because they carry no userinfo. Load the adapter in tests with
fresh mitmproxy stand-ins scoped to its import instead of leaving them
in sys.modules.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(capture): parse URL controls and mixed network authorities
* fix(capture): distinguish native paths and URL base authorities
* fix(capture): preserve file-relative path evidence after separators
* fix(comparison): recheck historical calls-only changes (#2223)
* fix(comparison): preserve incomplete function call evidence
Keep missing, unknown, partial and unresolved call relationships unknown in function comparisons and reconstruction claims. Aggregate IDA caller and callee coverage without overwriting uncertainty.
Validate with 144 focused tests, typecheck/lint, docs checks and test discovery. Real providers were not verified; the full headless IDA session suite requires unavailable Windows native controls.
* fix(comparison): recheck historical calls-only changes
---------
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
* fix(android): make DEX descriptor pattern portable
* fix(ci): isolate Linux Hopper process fixture inventory
* fix(capture): remove empty URL userinfo markers
* test(capture): cover empty URL userinfo markers
* test(ci): isolate Hopper process inventory fixtures
* fix(ci): declare process inventory callback
* test(ci): isolate alternate launcher process fixture
* fix(capture): keep URLs that WHATWG rejects instead of redacting userinfo
File URLs cannot carry userinfo (an "@" in a file host fails the parse),
and a special authority whose host or port WHATWG rejects fails the whole
URL. Neither has credentials, so the bridge now keeps their source text
and byte sidecar. Checked against Node's URL over 30,000 spellings: no
under-redaction, over-redaction 20,540 -> 5,865 (remaining: empty
userinfo, IPv4 range, IDNA, non-special and baseless references).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(capture): strip trailing URL whitespace and validate IP hosts and ports
The previous host check counted trailing C0 controls and spaces, which
WHATWG strips, so `https://user:pw@host ` kept its password. Trailing
C0/space is now removed from the parse view (raw offsets are unchanged).
Special hosts now also follow the WHATWG IPv4 parser (ends-in-a-number,
hex/octal parts, range) and IPv6 literal syntax (zone IDs rejected), and
ports are bounded lexically before int(), which raised ValueError past
4300 digits on Python 3.11+ and failed the whole capture.
Node URL differential over 126,000 spellings: under-redaction 8,084 -> 0,
ValueError 2,550 -> 0; remaining special-scheme over-redaction is IDNA
and the pinned empty-userinfo case. Cost: +0.6 us per URL.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(capture): redact userinfo before hosts that need Unicode mapping
A non-ASCII special host may map under UTS #46 to an accepted host
(fullwidth digits, ideographic full stops, ignored code points), so
treat it as accepted instead of running the ASCII IPv4 check on it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
Co-authored-by: Lucien <164857476+dakjdakd@users.noreply.github.com>
Restore source-position-aware ambient aliases, scalar env/argv reads, candidate
endpoint worker data, Chrome DOMSnapshot empty strings, and accumulated Ghidra
LaunchSupport diagnostics. Align regression fixtures with declared current
contracts, while preserving identity, provenance, cancellation, and cleanup
checks. Assert workflow caveats independently at both established locations.
Validation: original five source failures reproduced on main; 307 focused tests
passed with one platform skip, 22 fixture follow-up tests and 22 workflow/MCP
tests passed; full static and docs checks passed. Real public http-server CLI
analyses preserve identical artifact identity, statistics, and 581-token selected
output. Real Chromium confirms restored DOM form destinations; isolated real
service-worker verification passes on main and the change. The complete local
browser run retained an unclassified later marker timeout; current-head CI
provides complete real-browser and recovery lane results. Native IDA/Ghidra
execution and complete local suite execution are not claimed.
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
Co-authored-by: N0zoM1z0 <161784452+N0zoM1z0@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(uninstall): remove the REA-owned ZCode fallback registration
ZCode loads the user-scope `~/.agents/mcp.json` `mcpServers` fallback
whenever its native `mcp.servers` table is empty. Uninstall removed REA
only from the native table, so a REA-owned fallback entry became eligible
to load again once the last native server was gone, while uninstall
reported ZCode as removed and complete.
Also visit the ZCode fallback as a standard `mcpServers` JSON
configuration in the uninstall host. It goes through the same
ownership check, `.rea.backup`, readback and restore path as every other
client configuration, so other fallback servers and settings, and a
foreign `rea` entry, are preserved. Its result merges into the ZCode item.
Fixes#2664
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(uninstall): preserve the strict ZCode fallback dialect
Address review findings at their producing boundaries and preserve directly affected consumers. Source inspection and git diff --check completed. Tests and automated checks were not run at the user’s explicit request.
* fix(comparison): recheck historical calls-only changes (#2223)
* fix(comparison): preserve incomplete function call evidence
Keep missing, unknown, partial and unresolved call relationships unknown in function comparisons and reconstruction claims. Aggregate IDA caller and callee coverage without overwriting uncertainty.
Validate with 144 focused tests, typecheck/lint, docs checks and test discovery. Real providers were not verified; the full headless IDA session suite requires unavailable Windows native controls.
* fix(comparison): recheck historical calls-only changes
---------
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
* fix(ci): isolate Linux Hopper process fixture inventory
* fix(android): make DEX descriptor pattern portable
* fix: repair the static-check failures on the default branch (#2895)
* fix: repair the static-check failures on the default branch
The default branch fails its own required `Static checks` job three ways, so
every pull request reports that job as failed regardless of its own change.
- `npm run format:check` reports `src/domain/functionComparison.test.ts` and
`src/ida/IdaOperationRunner.ts`.
- `npm run lint` reports a duplicate `import { join } from "node:path"` in
`tests/boundary/mcp/ghidraInventoryMcp.test.ts` (lines 4 and 17).
- `npm run typecheck` reports `format: "devin"` missing from the client format
union in `src/application/SupportedClients.ts`, and an unchecked index in
`tests/boundary/cli/clientProfileLifecycle.test.ts`, whose pair array is
inferred as `string[][]`.
Apply the pinned formatter, drop the duplicate import, list the `"devin"` format
that its own definition already uses, and mark the pair array `as const`.
Validation: `npm run check:fast` and `npm run format:check` pass.
* fix(setup): keep Devin on the JSON configuration dialect
---------
Co-authored-by: 吴佳浩 <wujiahao@supremind.com>
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
Co-authored-by: Lucien <164857476+dakjdakd@users.noreply.github.com>
Co-authored-by: 吴佳浩Alben <35291372+XiaoBinGan@users.noreply.github.com>
Co-authored-by: 吴佳浩 <wujiahao@supremind.com>
* fix(reference): report entries beyond the path limit as I/O failures
Reference-source traversal joins each child onto the root path. A valid
tree whose nested entries exceed the platform's absolute path limit
(PATH_MAX 1024 on macOS) makes lstat fail with ENAMETOOLONG. The
reader's filesystem error allow-list lacked that code, so
filesystemFailureDetail returned undefined and the native rejection
escaped both readReferenceSource and importReferenceSource instead of
becoming the existing failed `io` entry.
Add ENAMETOOLONG to the allow-list, matching the JavaScript application
reader's list. Readable siblings stay in the graph and the unreachable
entry is recorded with its OS message.
Fixes#2652
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(reference): assert the imported graph keeps readable and failed entries
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(process): report no remaining time after a provider deadline fires
Node timers can fire while performance.now() is still a fraction of a
millisecond short of the monotonic deadline. ProviderStartupDeadline then
reported interruption "timeout" and an aborted signal while remainingMs()
returned ceil(0.x) = 1, so polling loops guarded by remainingMs() > 0
(Hopper and Ghidra startup) could run past an elapsed deadline, and the
real-timer test "aborts a bounded wait when its real monotonic deadline
expires" failed intermittently in CI (expected 1 to be 0).
Once the deadline's own timer has classified a timeout, report zero
remaining time. A deterministic regression test fires the timer with the
monotonic clock 0.4 ms short of the deadline.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(reference): preserve path limit entry uncertainty
* fix(reference): preserve unknown entry failures and identity
* fix(reference): type unknown-kind exclusions and keep them out of symlink counts
The metadata-failure path offers "unknown" to exclusion predicates, but
the predicate type only admitted concrete kinds, which broke the build.
Directory-only secret and ignore rules now also match an unknown-kind
entry (it may be a directory), and the summary counts unknown entries
separately instead of as symlinks.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
Co-authored-by: Codex <codex@users.noreply.github.com>
* fix(native): bound captured output carried by native command failures
The native command runner retains up to 64 MiB of output because
successful analysis needs it complete, and failureCapture copied all of
it into the failure. An inspect_plist whose plutil output exceeded the
budget therefore returned a 67 MB resource_constraint error whose
details.captured_output.stdout repeated the whole retained output.
Keep a 64 KiB UTF-8-safe excerpt per stream in failure diagnostics,
mark them truncated, and keep the observed stdout/stderr byte totals.
Successful captures are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(native): bound codesign failures and cover excerpt cuts
translateCodeSignExitFailure wrapped an accepted nonzero codesign capture
into a NativeCommandFailure by copying both complete streams, so
inspect_signature failures still bypassed the diagnostic bound. Route it
through the shared failureCapture builder; accepted output stays
complete until the provider decides it is a failure.
Add a focused runner test for an ordinary nonzero exit with stdout and
stderr just above the excerpt size and a two-byte character straddling
the cut, and a provider test for the codesign path.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(native): build excerpt test output inside the child process
The output strings embedded in one node -e argument totalled about 133
KB, above Linux's 128 KiB per-argument limit, so spawn failed with E2BIG
on the CI runner. Generate the repeated output in the child instead.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(process): order interaction differences by their dispatch time
first_divergence ranks candidate differences by the earliest recorded
time of the differing observations, but eventTime read only `at_ms`.
Interaction events record their time as `dispatched_at_ms`
(and `scheduled_at_ms`), so a changed interaction always ranked as
untimed. An interaction outcome that differed at 6 ms therefore lost to
a process sample that differed at 500 ms.
Give each comparison dimension its recorded time field. Interactions use
the dispatch time, when their outcome was observed; the other
dimensions keep `at_ms`. An earlier process difference still wins.
Fixes#2645
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(process): require ordered interaction dispatch times in captures
Comparison ranks a changed interaction by its dispatch time, while
firstCollectionDivergence returns the first differing interaction by
sequence. The capture parser checked only that interaction sequences are
contiguous, so an imported capture whose dispatch times decrease could
make that first-by-sequence difference later than another differing
interaction.
Interactions are dispatched in sequence and their times come from the
monotonic capture clock, so require non-decreasing dispatch times at
parsing, as frames, rendered frames, process samples and filesystem
checkpoints already require for their timestamps.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Node timers can fire while performance.now() is still a fraction of a
millisecond short of the monotonic deadline. ProviderStartupDeadline then
reported interruption "timeout" and an aborted signal while remainingMs()
returned ceil(0.x) = 1, so polling loops guarded by remainingMs() > 0
(Hopper and Ghidra startup) could run past an elapsed deadline, and the
real-timer test "aborts a bounded wait when its real monotonic deadline
expires" failed intermittently in CI (expected 1 to be 0).
Once the deadline's own timer has classified a timeout, report zero
remaining time. A deterministic regression test fires the timer with the
monotonic clock 0.4 ms short of the deadline.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
observe-native-ui and capture-native-ui-scenario were registered in one
loop with a shared option schema and description. The passive command
therefore advertised `--steps` ("Observe or run a scenario"), accepted
any value, and then dispatched observe_native_ui without it, so a
caller who selected the wrong command never learned that no scenario ran.
Register each command against its own contract. Observation declares
only the passive scope options, so `--steps` is rejected as an unknown
flag; the scenario command keeps its JSON step decoding and validation.
Each command now has its own description.
Fixes#2674
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* perf(evidence): reuse sealed bundle reads in compare and export
`rea compare`, the compare_bundles MCP tool and `rea evidence-export`
read bundles through the mutable reader, so each consumer
re-authenticated every record again: compareBundles parsed both bundles
a second time and bundleDigest's serializeEvidenceBundle a third time;
export's writer parsed the source once more before serializing.
These consumers only read the bundle, so read it through the existing
sealed reader (readImmutableEvidenceBundle). Later parses hit the
immutable bundle cache. Ordinary mutable reads, kept deliberately by
#1721, are unchanged. Output is byte-identical.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(process): report no remaining time after a provider deadline fires
Node timers can fire while performance.now() is still a fraction of a
millisecond short of the monotonic deadline. ProviderStartupDeadline then
reported interruption "timeout" and an aborted signal while remainingMs()
returned ceil(0.x) = 1, so polling loops guarded by remainingMs() > 0
(Hopper and Ghidra startup) could run past an elapsed deadline, and the
real-timer test "aborts a bounded wait when its real monotonic deadline
expires" failed intermittently in CI (expected 1 to be 0).
Once the deadline's own timer has classified a timeout, report zero
remaining time. A deterministic regression test fires the timer with the
monotonic clock 0.4 ms short of the deadline.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
* fix(flutter): name a missing or non-file APK selection
identify_flutter_build and inspect_dart_aot reported a missing path or a
directory as "The APK could not be read as a package: <path>: ENOENT ...",
which blames the package bytes for a selection mistake. Report
"Selected APK does not exist" for ENOENT/ENOTDIR and "Selected APK is not
a regular file" for a non-regular selection, matching the SQLite
inspector's wording. The issue path and invalid_format reason are
unchanged; package format failures keep their existing message.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(filesystem): classify unopenable non-regular selections at admission
A Unix socket cannot be opened at all: open() fails with ENXIO on Linux
and EOPNOTSUPP on macOS (surfaced by libuv as "Unknown system error
-102") before openRegularFile can stat the descriptor, so such a
selection escaped NonRegularFileReadError and the Flutter provider still
blamed the package bytes.
When open fails, classify the path by its own metadata (lstat when
symlinks are rejected, stat otherwise): an existing non-regular file
becomes NonRegularFileReadError. Missing paths, permission failures,
regular files, and the O_NOFOLLOW ELOOP rejection keep their original
errors. The Flutter selection test now covers a real Unix socket.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(filesystem): reclassify only unopenable file-type errors at admission
Checking the path's metadata after any open failure turned EACCES on an
unreadable directory into NonRegularFileReadError, so callers reported
invalid input instead of access_denied. Reclassify only errors that mean
the file type cannot be opened (ENXIO, EOPNOTSUPP, ENOTSUP, matched by
errno because libuv has no name for macOS EOPNOTSUPP 102); permission,
symlink-rejection and resource failures keep their original errors.
Run the Unix socket selection check in its own test, skipped on Windows
where listen() takes named pipes, and keep the missing-path and
directory checks cross-platform.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(filesystem): keep cancellation that arrives during the non-regular probe
A caller abort while the post-open metadata probe ran was replaced by
NonRegularFileReadError, so callers such as parseBinaryTarget reported an
input-format error instead of cancellation. Recheck the signal before
substituting the non-regular classification.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
inspect_artifact parsed a fresh inventory and inspection report but left
both mutable, so each later Evidence boundary cloned and re-digested the
complete payload: createArtifactInspection re-authenticated the nested
inventory Evidence, createAnalysisExecution re-parsed the report, and
recordEvidence re-authenticated the outer Evidence.
Seal both exclusively owned values with freezeJsonSnapshot so those
boundaries take the existing immutable-snapshot path. Output is
byte-identical.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The decoder already reports "Instruction X at IL offset N leaves method
body" when an operand's computed end passes the IL body, but only after
the operand was read. Operand readers use Buffer reads, so an operand
cut off by the body end threw a RangeError first. The catch-all stored
Node's message ("Attempt to access memory outside buffer bounds" or an
offset RangeError) as body.issue, dropping the instruction and offset.
A body ending in a lone 0xFE two-byte opcode prefix failed the same way.
Treat a RangeError from the operand read as the operand leaving the body
so the existing diagnostic applies, and check the second opcode byte
before reading it. Malformed status, partial coverage and earlier
decoded instructions are unchanged.
Fixes#2677
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(artifacts): report malformed ZIP archives as format failures
zip.js signals archive faults (missing end-of-central-directory record,
unrecognized format, overlapping members, corrupt member data,
unsupported compression) as plain Errors. ZipArtifactReader passed them
through untyped, so the artifact provider's fallback reported every one
as `io` with no detail and told the caller to "check file access".
Type these faults at the ZIP reader boundary: structural and data
corruption become `format` failures and unsupported archive features
become `unavailable`, each carrying the zip.js reason as detail. This
covers enumeration, the getData rejection, and the WritableStream abort
path that delivers corrupt-data errors first.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(artifacts): share the ZIP inventory failure assertion
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Add a plain x86_64 slice to the dyld fixture while retaining both same-named arm64e slices and their independent environment assertions.
The original fixture was rejected by real x64 CLI admission before tracing. The contributing PR's complete Linux CI run passes all 30 jobs, including the portable suite. Codex and Greptile reviews completed without findings; no unresolved threads remain.
This restores the main test baseline needed for #2635. The change is limited to test data and preserves the original author's contribution.
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
deriveFindings looked up each traced root's slice again by architecture
name. A universal executable with arm64e and arm64e.x1 slices reports
both as "arm64e", so both roots resolved to the first slice: a later
slice's LC_DYLD_ENVIRONMENT produced no finding, or the first slice's
command was reported twice. The per-slice dyld_environment arrays were
already correct.
traceDylibLoading already has the exact slice when it records a root, so
pass it to deriveFindings instead of re-deriving it from the name. The
public roots and findings shapes are unchanged.
Fixes#2661
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
HAR validation reported only the AJV keyword, so a capture missing
`log.version` failed with "HAR schema validation failed: required." and
the caller could not tell which property to supply. Include AJV's
constraint message, which names the missing property, expected type, or
pattern without restating captured values.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* perf(native): add opt-in Mach-O summary delivery and retained pages
inspect_macho returns every import and export inline. For VS Code's
Electron Framework one MCP call carries 2.5 MB of text (about 828k o200k
tokens), and the same again as structuredContent; the MCP response
budget admits it. Agents asking for slice identity, UUID or dependency
counts paid that cost on every first call.
Add the established opt-in detail selection: omitted or complete keeps
the existing complete result and Evidence identity, while summary
retains the complete Evidence and returns slice identity with
per-collection returned/total/exhaustive counts (956 tokens).
inspect_analysis_view kind macho pages the original rows of any
collection or the command provenance, repeating the producer's own
total and exhaustive flag so page exhaustion is never read as absence.
Delivery detail never reaches the native tools or the source Evidence
ID: registerEvidenceTools gains a parametersForEvidence hook, so the
summary's parent is the same Evidence as a complete call.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(mcp): refuse unretainable summary delivery before analysis
A server without Evidence retention (a fixed analysis source) ran the
complete inspect_macho or inspect_artifact analysis for detail summary
and only then refused it in summarizeRetainedAnalysis, paying the very
cost summary delivery exists to avoid. registerEvidenceTools gains a
preflight hook; both registrations check retention there, as the WASM,
EVM, managed, pcap and capture handlers already do before dispatch.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(native): keep whole rows in Mach-O architecture and segment pages
The page item union listed plain-object row arrays, and zod returns the
first union member that parses while dropping unknown keys. Entrypoint
rows declare only file_offset, so every architecture and segment row
(which also carry a numeric file_offset) matched it and lost its name,
CPU fields, permissions and sections. Rows now use exact key sets, and
imports/exports share one member. The MCP test pages every facet of a
real /bin/ls and compares it with the unparsed complete rows.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Once a Frida session reached its 16 MiB capture budget, every new message
evicted the oldest one with Array#shift, re-serialized it to recover its
byte charge, and searched and spliced its per-script array. Each of
those is linear in the retained messages, so with about 188k small
messages retained the cost per message rose from 0.14 µs to 32 µs and
capped the main thread at roughly 31k messages per second, which a
hooked hot function can exceed.
Keep both queues as head-offset FIFOs that compact once half their slots
are evicted, and record each message's charge when it is admitted.
Arrival order already makes the globally oldest message the oldest one
of its script, so per-script eviction becomes a head check. Eviction is
now amortized O(1) at about 0.14 µs per message; retained messages,
their order and the truncation flag are unchanged.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(native): name the ownership helper prerequisite in command failures
Every supervised native command launch needs the Darwin process-ownership
helper. When that helper cannot be compiled (relative REA_DARWIN_SWIFTC_PATH,
missing SDKROOT, a missing compiler, or an SDK/compiler mismatch), the
runner reduced the typed DarwinProcessOwnershipInspectionError to a bare
`io` failure. inspect_macho then reported `execution_failure` with no hint,
and inspect_artifact on any universal Mach-O said only "lipo could not
enumerate universal Mach-O slices".
The runner now keeps the inspection message as the failure's `detail` at
the launch boundary and carries it through xcrun resolution. The native
provider adds it to its diagnostics and the slice reader appends it to its
message, matching what the SQLite and capture providers already report.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(native): cover the provider's ownership helper diagnostics
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The pcap summary sorted equal-count endpoints and protocol stacks with
String#localeCompare, and Flutter build inspection sorted APK ABI
directory names the same way. localeCompare follows the host's ICU
locale, so the same capture produced differently ordered Evidence on
different hosts: under da_DK "aarp" sorts after "zz", under lt_LT
"syslog" sorts before "ssdp", and every locale puts "fe80::1" before
"fe80:1::1". Use the repository's compareUnicodeCodePoints, as the
other Evidence producers already do.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(pcap): name the host tool to configure in capability refusals
Missing or misconfigured tshark/capinfos, the unsupported macOS capture
boundary and an unusable Linux prlimit were projected with the generic
"unavailable for the current target; choose another target" advice, even
for inspect_pcap_client, which has no target. Route tool refusals through
one pcapToolUnavailable owner that names REA_TSHARK_COMMAND and
REA_CAPINFOS_COMMAND, give host-boundary refusals their own reason, and
keep the private workspace refusal's retryable flag like WABT and source
maps do.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(pcap): cover the Linux prlimit refusal guidance
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
On hosts without an enforceable native bound (macOS), the provider refuses
inspect_pcap_capture with capability_unavailable before it touches the
path, which the unsupported-boundary test in the same file pins. The
missing-path test expected invalid_request on every host, so it always
failed on macOS while Linux and Windows CI stayed green.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
inspect_macho launched otool, nm, two dyld_info reads, dwarfdump and
vtool one at a time. Each launch also resolves the tool through
`xcrun --find`, so a warm call spent about 350 ms in eight serial
supervised launches. The slice commands only read the selected
artifact, so launch them together. Every launch still settles before
the result, and declared order still fixes provenance and which
failure is reported.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
When a pwntools or pwndbg stage cannot confirm that its owned process
group stopped, the cleanup_incomplete result listed only the run ID and
runtime directory, although a descendant can still be live. Add the
retained supervisor's pid and process-group coordinates, as
OwnedCommand reports for a non-transferred cleanup failure.
Fixes#2274
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(frida): reject an unreadable script file before starting a session
The shared instrument_with_frida workflow started the Frida session, which
attaches or spawns the target, before the provider read a file script
source. A missing script file then failed loadScript with a bare
invalid_request, and session cleanup resumed the paused spawn, so the
target ran without instrumentation. Check that the selected file is a
readable regular file first and report the path, or access_denied for
EACCES/EPERM, before any device effect.
Fixes#2209
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(frida): let cancellation win over the script file check
An already-aborted request with an unreadable file source returned the
input or access-denied error instead of AnalysisCancelledError. Check the
signal before the script file preflight and again after it, so file
sources keep the existing cancellation semantics.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* refactor(frida): keep the pre-start checks out of instrument
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Frida session lookups, script lookups, unreadable script files and the
instrument-with-frida PID and script checks returned invalid_request with
no issues, while the remediation told callers to correct the listed
arguments. Attach the session_id, script_id, path, target or script issue
with the reason so CLI and MCP callers can see what to fix.
Fixes#2213
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
startMcpTransport loads optional observation providers once per process
and passes them to every server instance the SDK constructs. Each
instance's closeProviders also closed the shared browser scenario
provider, so when the SDK discarded a server/discover probe instance for a
legacy initialize, every later capture_browser_scenario call failed with
"Browser scenario provider is closing." Server instances no longer close
the caller-supplied provider; the transport returns
closeOptionalProviders and process shutdown closes it once.
Fixes#2195
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(setup): remove emptied skill directories on uninstall
Uninstall removes only the managed skill files so user files survive,
but it left every emptied reverse-engineer-anything directory and its
references/ subdirectory behind. After the ownership manifest is removed,
remove the managed directories deepest first with a non-recursive rmdir,
which keeps any directory that still holds user files.
Fixes#2159
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(setup): report emptied skill directories that cannot be removed
Only a directory that still holds user files, is already absent, or is no
longer a directory is kept silently. Any other rmdir failure, such as a
parent that denies deletion, now fails the skill item with the directory
path instead of reporting the skill as removed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
A client configuration that parses but whose root or MCP server table is
not an object (for example {"mcpServers": null} or []) made setup print a
raw Zod issue dump and made uninstall claim the file was not valid JSON.
parseClientConfiguration now raises a ClientConfigurationShapeError naming
the root or the servers key, which setup reports as the malformed detail
and uninstall reports instead of a syntax error.
Fixes#2152
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(managed): report malformed PE layouts as invalid input
readManagedPeLayout raises ManagedReaderFailure for DOS, COFF and section
header failures such as section data past EOF. Those escaped the managed
inspectors, so ManagedStaticProvider wrapped them as ProviderAdapterError
and the CLI and MCP reported execution_failure with advice to run rea
doctor. Map format failures to an invalid_format input issue carrying the
reader's detail, as inspect-pe-resources and inspect-go-binary do.
Fixes#2191
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(managed): route unmappable PE layouts through one input error
inspect-managed-native-boundaries caught a layout-level ManagedReaderFailure
itself and returned an empty malformed inspection with exit 0, while the
other managed operations now report the same bytes as invalid input. Let
the layout failure reach the provider's invalid_format mapping; metadata
failures after a readable layout still return malformed coverage.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
JebProvider throws when REA_JEB_MCP_URL fails endpoint validation, and the
composition built it eagerly for the CLI and every MCP connection. A
malformed value, or http://localhost:8425/mcp, therefore crashed every CLI
command with a stack trace and failed every MCP connection with an
internal server error. The composition now parses the endpoint first and,
when it is unusable, returns a JEB port that reports not_configured
availability and a ConfigurationError naming REA_JEB_MCP_URL.
Fixes#2143
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
createCli parsed REA_LOG_LEVEL with a throwing schema, so an invalid or
uppercase value crashed every command, including doctor and --help, with
a ZodError stack trace. Parse it into a ConfigurationError result and have
the executable print the same configuration message as the MCP adapter
and exit 1 before building the CLI.
Fixes#2135
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* ci: narrow native-host lane selection and reuse build caches
Unowned src/ and bridge/ files no longer fall back to the full baseline.
They run the Linux suite plus the Linux x64 installed-package/Inspector
baseline. New apple and package scopes own the Apple artifact lane, the
package matrix and the Windows curated lane, including the test files those
lanes invoke. Manual-only verifier entrypoints run static checks and the
Linux suite. CHANGELOG, SECURITY, GitHub templates and the native Windows
README are documentation-only.
Pull requests omit the macOS x64 package host unless labelled ci:full.
The Windows and Apple lanes restore the shared Turbo build cache, the
Windows lane drops its duplicate typecheck, and the Linux suite uses six
shards. Main pushes keep the full baseline.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* ci: match package script entrypoints against scope globs
Script-only package.json edits resolved their entrypoint with literal
equality, so glob-owned verifiers such as scripts/verify/apple/** and
scripts/verify/inspector/** fell back to the full baseline instead of
selecting their owning lane.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Uninstall checks every skill destination but labeled each result only as
"skill" (or "Claude Code skill"), so a retained or failed destination
could not be located. Label each result with the directory it refers to.
Fixes#2111
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
`inspect-managed-members --type` accepted repeated and empty names that
the MCP contract rejects. The provider's contract parse then threw, and
the CLI reported execution_failure with advice to run `rea doctor`.
Reuse the contract's type_names schema so the CLI rejects them as a
validation error on --type.
Fixes#2121
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Target resolution admitted a property list only when its name ended in
.plist, so keyed archives with other extensions (for example .archive)
were rejected as an unsupported binary format before any plist tool ran.
Classify bplist00 bytes as a plist, as the artifact inventory already does.
Fixes#2129
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
07b36c148 made document and procedure completion read only from an
already-started provider client. Two checks still completed right after
opening a target, so they observed an empty cold result:
- promptCompletion.test.ts used a fixture client without executeIfReady
and never ran an ordinary operation, so the composite provider had not
created a client yet.
- The packaged MCP smoke expected live completions on non-Linux hosts
immediately after open_binary, before current_document started Hopper.
Both now assert the cold result is empty, ready the provider through an
ordinary operation, and then assert the unchanged live candidates.
Fixes#2123
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Update the star milestone in the English README and all 18 translated
READMEs, keeping each locale's number formatting.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Claude Code turns MCP tool search off when ANTHROPIC_BASE_URL points to a
non-first-party host, so every REA tool definition is sent with every
request. On main at 04089d5be the forwarded names, descriptions and input
schemas measure about 496 KB (~120k tokens) with the full profile and about
290 KB (~68k tokens) with the compact profile. When Claude Code assumes a
smaller window than the model has, as with `deepseek-v4-flash` without the
`[1m]` suffix, autocompact thrashes because compaction cannot remove tool
definitions.
Document the window, tool-search and compact-profile remedies, and that
DeepSeek's Anthropic API does not list `tool_reference` blocks.
Refs #2019
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Profile resolution gives Ghidra and Hopper a concrete version, but the
selected candidate's capabilities and the session route kept the
provider's static identity with a null version. binary_session therefore
grouped capability sets under a different identity than the binding and
the selected candidate.
The registry now rebinds the selected provider's capabilities to the
resolved identity once, on the binding, and the route uses them. Evidence
provenance already compensated downstream by substituting the profile
identity; that repair is removed because the route now carries the
resolved identity.
Fixes#2009
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(native): reject required otool fields truncated before their value
#1967 let requireUniqueField count a label at end of line so an empty
dylib-use options row is present. That applied to every required field,
so truncated output such as a bare vmaddr, vmsize, uuid or entryoff line
passed the presence check while parseField dropped it, and the value
silently became null.
Let parseField decide which labeled line carries a value: the single
matching line must parse as that field. The empty options row remains
the only valueless field parseField admits.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(native): admit the empty segname of Mach-O object files
MH_OBJECT files put their sections in one segment whose name is empty,
and otool -l prints it as a bare segname label. Requiring a value for
every scalar field rejected valid object-file output as unreadable.
Let parseField admit an empty segname like the empty dylib-use options
row, and keep projecting an empty or absent segment name as "unknown",
as main did.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(native): admit all-NUL Mach-O section names
An all-NUL section_64.sectname prints as a bare sectname label, like an
empty segname. Admit it in parseField and keep projecting the empty name
as "unknown", as main did.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The ADB provider, which the apktool provider mirrors, had the same two
defects: a relative REA_ADB_PATH escaped as UNKNOWN because selection
ran before the try that maps configuration failures, and a missing adb
binary projected generic target advice instead of its reason.
Move the configuration-failure-to-refusal mapping into one
providerConfigurationRefusal helper used by both providers, resolve the
ADB selection inside the mapped try, and share the CLI refusal assertion.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
Inventory attribute lookups and the members inspector each interpreted
TypeDef member-list ownership. The members lookup binary-searched ordered
ranges but fell back to a per-member scan of every type when malformed
lists were unordered, which was quadratic (8 s for 100,000 types).
Both now use firstMatchListOwners, one near-linear first-match index over
one-based list positions. Remove the ordered/fallback split and the
production declaringType scan, which remains only as the test oracle. The
inventory reserves only its retained index.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
A missing apktool launcher projected the generic "unavailable for the
current target" message and remediation, leaving the actual reason only
in details.reason. A relative REA_APKTOOL_COMMAND escaped as UNKNOWN
because command selection ran before the try that maps configuration
failures.
Resolve the selection inside that try for both execution and
availability, and pass the refusal reason as userMessage, as JADX and
the adapters fixed in #1224 do.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Each CustomAttribute whose constructor is a local MethodDef scanned the
TypeDef table to find the declaring type, so inspect_managed_artifact did
O(attributes x types) row decoding. A 584 KB assembly with 16,000 types
and 16,000 such attributes blocked for about 38 s.
Build a first-match owner index once per metadata layout. Already-owned
list positions are skipped, so overlapping malformed method lists keep
their previous first-match owner in near-linear time. The index is
charged to the decode budget, and a build failure is cached like the
member pointer lookup.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): size the future static reserve by remaining sources
The parent reserved 80% of its result budget for future static facts
whenever any source remained, so one tiny later file could discard an
earlier file's completed semantic graph. Reserve at most one bounded
static checkpoint's retained size per remaining source, within the same
shared pool, so many remaining sources keep the previous protection.
Parent-budget rejections now report the effective admission budget and
explain that the CLI/MCP process heap, not max_heap_mb, bounds them.
Refs #1911
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): make parent budget remediation and its test runtime-aware
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): suggest a concrete parent heap size in remediation
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): let parent budget rejections state their retained facts
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(errors): stop recommending retries for non-retryable failures
Unreadable output and provider adapter failures project retryable=false,
but their standard message and remediation said "Retry once", so a
firmware run whose expected report was never published told callers to
repeat a non-retryable operation. Point those failures at `rea doctor`
and the reported details instead, leaving retryability decided once by
the projected error code.
Fixes#1855
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(errors): align Hopper remote failure guidance with retryability
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(managed): decode generic, array, fnptr, and modifier signatures
Member signatures using GENERICINST, ARRAY, FNPTR, CMOD_REQD/CMOD_OPT,
or TYPEDBYREF were reported as unsupported, leaving their types unknown
and turning otherwise complete inspections partial. Decode them with an
explicit stack so nested signatures stay recursion-free, keep encoded
TypeDefOrRefOrSpec values, and report malformed shapes as malformed.
Fixes#1832
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(managed): bound array rank before rendering dimensions
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(managed): keep modifier order and reject non-method fnptr headers
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(managed): reject corrupt fnptr flags and misplaced typed references
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(managed): decode signature elements by grammar position
Track whether each element is a field type, return type, parameter,
nested Type, or pointee. VOID is valid only as a return type or
pointee, TYPEDBYREF only as a return type or parameter, and BYREF only
at the top of a return type, parameter, or ref field; other placements
are malformed. Function pointer items are parsed as a return type then
parameters, and generic arguments as Types.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
When partial inventories omitted node details, the root got an unknown
availability row and the differing manifest root digests appended a
second, changed content row, so one logical path was counted as both
changed and unknown. Fold the manifest content change into an existing
unknown or changed root row instead, keeping its dimensions, occurrence
and artifact identities, and citations; contradiction and presence rows
are left as they are.
Fixes#1844
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Update the star milestone in the English README and all 18 translated
READMEs, keeping each locale's number formatting.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(managed): report ExportedType rows in managed member inspection
Facade assemblies such as System.Runtime.dll provide their types only
through ExportedType rows (type forwarders), which member inspection
never read, so they reported complete coverage with no types. Report
each row as declared with its flags, forwarder bit, TypeDefId hint, and
implementation token, and record invalid or null implementation coded
indexes as partial coverage. Comparisons now state that exported types
are not compared when either side has them.
Fixes#1812
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(managed): read earlier member Evidence without exported types
Member Evidence recorded before ExportedType rows were read omits the
field. Parse it as null (not examined) instead of rejecting it, and
state in comparisons that such an observation's exported types are
unknown.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(managed): leave exported types unknown when metadata is unavailable
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): follow bundler namespace interop for Electron APIs
Rolldown/tsdown emit `let electron = require("electron"); electron =
__toESM(electron);` and esbuild/tsc emit `__toESM(require("electron"))`
or `__importStar(require("electron"))` for namespace imports. The
reassignment or the helper call hid the Electron origin, so bundled
main processes reported zero BrowserWindows and preloads.
Locally declared, never reassigned interop helpers recognized by name
now pass the module through, and a binding's rewrap of its own value is
not treated as a reassignment. Other writes stay unresolved.
Fixes#1802
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): require emitted interop helper definitions
Only trust an interop helper name when it has a single program-scope
function definition, and state that helper bodies are not evaluated.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Windows has no POSIX execute bit; Node reports file modes from the
read-only attribute alone, so every host file, including PE images,
reported `executable: false` while the same occurrence reported
`artifact_kind: executable`. Derive the host executable marking once:
POSIX permission bits stay authoritative, and Windows marks regular
files with the CreateProcess/cmd launchable extensions.
Fixes#1819
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
On Windows, Node refuses to spawn .bat and .cmd files without a shell
(CVE-2024-27980), so REA_APKTOOL_COMMAND pointing at Chocolatey's
apktool.bat or a java -jar wrapper failed its version probe with EINVAL,
reported as "did not report a usable version".
Move Ghidra's conservative cmd.exe invocation (no AutoRun, no delayed
expansion, every token quoted, interpreter metacharacters rejected) into
src/process/WindowsBatchCommand.ts. Ghidra keeps its error type, wording
and length limit. apktool now uses it for both the version probe and the
decode run when the launcher is a batch file on Windows. A launcher that
cmd.exe would reinterpret reports launcher_unsupported (not_configured)
with remediation instead of version_unresolved.
Fixes#1774
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(hopper): let failed one-shot commands exit while retaining a lease
A Hopper target lease retained after unconfirmed cleanup kept its
listening socket referenced, so a one-shot CLI emitted its final
cleanup_incomplete result but never exited, and its live lease rejected
every retry for the same target as already open.
Unreference the lease listener. A live MCP process still holds the
lease for as long as it runs; a completed CLI exits with status 1, and
the stale socket it leaves is recovered by the next acquirer.
Fixes#1751
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(hopper): remove only this test's stale lease sockets
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Attached IDA mode always uses the legacy 1.4 producer, whose
disassemble_function jumps to the function and whose decompile_function
opens or reuses a pseudocode view whenever an IDA window is active.
Session discovery still advertised may_show_ui: false for every
operation.
Declare the UI effect for the attached operations that reach those
tools: pseudocode, assembly, raw function instructions and the function
dossier. Headless operations and the other attached operations keep
false. A conformance check now compares each declaration with the
upstream tools the operation actually calls.
Fixes#1712
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Section headers distinguish the declared virtual extent from the on-disk raw extent, which may include alignment padding. Resource RVA ownership and overlap use VirtualSize, falling back to SizeOfRawData when VirtualSize is zero, and payload reads still require file-backed bytes.
The NativeAOT snapshot reader applies the same range rule to RVA reads, metadata scans, executable targets, and derived overlay bounds. Page-rounded loader mappings may expose tail bytes, but those bytes lie outside the declared section range used by these parsers.
analyze_javascript_application ran several long phases with no event-loop
turn, so MCP ping and cancellation waited: each file's parse, analysis and
semantic projection; building every application node and relationship
before the application graph's first sealing step; and whole-graph schema
parses inside both immutable graph factories.
- Each file runs as steps (parse, static analysis, semantic analysis, then
each projection pass), driven by the existing time-sliced checkpoint.
- Application graph input construction yields between phases, per module
link and per file of findings.
- The pipeline transfers its privately built graph inputs to new sealing
steps that validate the graph's own fields synchronously and elements,
content checks and the digest in steps. The public copying factories
keep their synchronous capture.
Results are unchanged: graph IDs and the complete result digest match main
on five inputs including Obsidian's 23 MB obsidian.asar. On the issue's
1,200-function fixture the longest event-loop block falls from about
290 ms to 90 ms, and ping after abort from up to 286 ms to at most 56 ms.
On obsidian.asar it falls from 10.9 s to 5.4 s; the rest is one analyzer
stage of a single very large file.
Refs #1462
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: N0zoM1z0 <161784452+N0zoM1z0@users.noreply.github.com>
Since #1471, escaping an object also escapes every method's results. Each
call through the object escapes it as the receiver, and each escape
re-expanded all of its methods. N methods returning one shared object,
each called once, did N times N work: 2,000 calls took 3.0 s of semantic
analysis and 4,000 took 12 s.
Expansion below references depends only on syntax, so escapes now reuse
it per expression and path. Identifiers that resolve to the same binding
and path collapse to one, because the escape walk treats them alike.
Every escape site is still recorded. The semantic IR is byte-identical
on 50 sources, and the issue's case takes 253 ms at 2,000 and 638 ms at
4,000.
Closes#1495
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
inspect_keyed_archive passed caller-selected binary plists straight to
plist.parseBinary, which recurses on every object reference and copies
shared containers. A 484-byte archive of forty nested [next, next] arrays
grew the process past 4 GB until it died with no result, which would also
end an MCP server. A 125-byte self-referencing array overflowed the stack
and was reported as an artifact too large to process.
The iterative reference walk that already guards decode_interface_builder
now runs first, with an 8x keyed-archive decode budget. Over-budget depth
or expansion is a limit; a reference cycle is a malformed archive. Its
"decoder" bounds admit the same layouts plist.parseBinary reads, so
accepted archives stay accepted. Interface Builder behavior is unchanged.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Strict JSON file inputs reject a leading UTF-8 byte-order mark, but the
diagnostic was the parser's own message, which quotes the invisible U+FEFF
as the unexpected token: Unexpected token '', "{"invento"... is not
valid JSON. Editors such as Windows PowerShell 5.1 write that mark by
default, and nothing in the message showed it.
CLI JSON files (whole and streamed), Evidence and snapshot JSON files, and
process scenario files now report that the input begins with a UTF-8
byte-order mark and how to save it. Acceptance is unchanged.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
HAR 1.2 lets writers emit a UTF-8 byte-order mark and requires readers to
ignore it. The decoder passed ignoreBOM: true to TextDecoder, which keeps
the mark in the decoded text, so a BOM-prefixed export failed as "HAR is
malformed JSON." Decoding now uses the default, which skips one leading
mark. The artifact digest still covers the selected file bytes.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
inspect_analysis_view is the recovery path when an analyze_function
dossier exceeds the MCP response budget, but its native facets covered
only procedure, pseudocode, assembly, callers/callees, references, and
value flow. Basic blocks, comments, unresolved calls, referenced strings
and names, and the native API record could not be read from retained
Evidence at all; basic blocks grow with the function just like assembly.
These are now facets: collections page by rows, and native_api is a
singleton. A dossier without an unresolved_calls record reports "not
recorded" rather than the value-flow failure reason.
Refs #1247 and #1065.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Inventory paired every path under one case spelling with every path under
each other spelling and listed all of them on each occurrence. An archive
with Foo/ and foo/ subtrees of N files each produced N-entry lists on 2N
occurrences: 4,000 such ZIP members yielded a 120 MB inspect_artifact
result and close to 1 GB of memory. The trie walks also recursed once per
path component, so a ZIP name with tens of thousands of components threw
RangeError.
A collision is now recorded at the segment where spellings diverge and
shared down that subtree by an iterative walk. Each occurrence names the
colliding prefix and the first other spelling, plus a count of further
spellings, each of which is its own occurrence.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Every inventory edge is created from one occurrence and copied that
occurrence's artifact_id and logical_path as child_artifact_id and
logical_path. Edges now keep the occurrence_id reference alone. Edge IDs
still hash the same child and path, read from the occurrence, so they are
unchanged; validation and artifact comparison read the occurrence too.
inspect_artifact output: Transmission.app 1.10 MB to 0.97 MB, Visual
Studio Code.app 15.37 MB to 13.48 MB.
BREAKING CHANGE: Artifact inventory edges no longer include
child_artifact_id or logical_path; read them from the occurrence named by
occurrence_id. The graph commitment changes, so inventory Evidence
recorded by earlier versions no longer validates.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
inventory_artifact, inspect_artifact, and extract_artifact copied each
occurrence or extracted file path into Evidence locations, although the
result already names every one. inspect_artifact did so twice, in its own
Evidence and in the nested inventory Evidence. Like other whole-target
Evidence, these now report no locations.
Visual Studio Code.app: inspect_artifact 15.37 MB to 13.07 MB,
extract_artifact 1.40 MB to 1.06 MB.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Input schema presentation gave every undescribed property a generic
"Value for <name>." sentence. That only restates the property name, yet
it covered 1,958 properties: 70 KB of the 522 KB of advertised input
schemas that clients load into model context. Authored descriptions,
referenced descriptions, and the name-pattern sentences that add meaning
(maximums, digests, paths, Evidence IDs) remain; other properties now
advertise no description.
The schema validity test now rejects placeholder descriptions instead of
requiring a description on every property.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The NIBArchive projection synthesizes customClass, label, and objectID so
its records share the ibtool dictionary parser. The parser then returned
those values as class_name, name, and the object ID, and the graph node
kept the synthesized copies in attributes as well. Attributes now omit a
key whose value its canonical field already carries; differing authored
values stay. Transmission.app decode_interface_builder output shrinks
from 6.45 MB to 6.14 MB.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Every semantic graph node and relation carried an identifier_strategy
object whose schema admits exactly one value per record type. The
constant made up about 9% of analyze_javascript_application output
(Obsidian app.asar: 12.04 MB to 10.96 MB), and semantic traces repeated
it on every returned fact.
The node_id and relation_id schema descriptions now state how each
identifier is derived.
BREAKING CHANGE: Semantic graph nodes and relations no longer include
identifier_strategy, and relation IDs no longer hash that constant.
Evidence recorded by earlier versions does not validate against the new
semantic graph schema.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
inspect_artifact nested the complete inventory Evidence and then restated
every artifact node and occurrence as an observation and every edge as a
derived relationship. No caller reads those copies; comparison and tests
use the nested inventory Evidence. On Transmission.app the result shrinks
from 2.46 MB to 1.10 MB.
Observations keep the root manifest and integrity findings.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Add an X link beside the GitHub and Discord header icons on every content
page, name the footer links "Join the community" and "Follow us on X", and
close the homepage with a "Join the community" section.
On narrow screens, pad the footer so the floating Top link no longer
covers its last row of links.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Every Interface Builder node and edge evidence record prefixed its
description with the archive's relative path, which artifact_path in the
same record already states. Bundles carry thousands of objects across
localized nibs, so the repeated path made up 6-10% of
decode_interface_builder output (XLD: 7.38 MB to 6.62 MB).
Descriptions now name only the decoded item, such as "object 81" or
"outlet o_quality".
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The semantic graph's evidence-context registry canonicalized the full
provenance context three times for every node, relation, unknown and
fingerprint: once to hash its ID and twice more to compare against the
stored row. A file has only a few distinct contexts, so nearly all of
that work repeated a known answer.
Interning now keys contexts by their canonical text. Each fact
canonicalizes once, and only a new context is hashed and checked for a
digest collision. IDs and output are unchanged; Obsidian's 334 MB
Evidence is byte-identical and analysis drops from about 47.0 s to
44.4 s.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Rendered frames padded every visible row with trailing spaces to the
terminal width. Blank rows cost a full row each and every frame repeats
them, so the padding made up 14-39% of capture Evidence (2 KB of sparse
output produced 228 KB). The padding carried no information: columns is
recorded and serialized_state keeps the exact terminal state.
Lines are now trimmed after normalization, so line.padEnd(columns)
restores the previous value. The comparison contract commits
rendered_line_format so older padded captures do not compare as the
same contract.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Load browser composition after CLI scenario validation and load Git support only for local reference trees with Git metadata. Preserve default outputs and actual command behavior.
Refs #1065.
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
The skill's CLI fallback printed the complete application Evidence to
stdout. A real ASAR produces hundreds of megabytes (334 MB for Obsidian),
which an agent then reads into its context. Save the Evidence to a file
and project a summary or module page with inspect-analysis-view instead.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
inspect_macho, inspect_signature, inspect_plist, list_architectures and
demangle_swift already return their command provenance inside the
normalized result. The provider also copied the same array into
raw_result, so every response carried it twice. raw_result now stays
null, as docs/mcp-contracts.md describes for results without a distinct
provider representation.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
`--json` / `--format json` always indented its streamed result. Agents and
scripts read REA through pipes and redirects, where indentation carries no
information: on real application results it added 16-45% model tokens
(o200k proxy) and 25-73% bytes. The skill's CLI fallback uses `--json`.
Indent only when stdout is a TTY, as `gh --json` does; piped or redirected
JSON is compact. The JSON value, `jsonl`, `--filter-output`,
`--full-output` and the token-window paths are unchanged. `jq .` restores
indentation for files.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
HopperProvider passed the bridge reply as rawResult and, for every
operation except address_to_file_offset, the same object as the result.
Every Hopper Evidence record (decompilation, procedure and string
listings, xrefs, ...) therefore carried its complete result twice, and
an MCP response four times counting its text copy.
raw_result is now null unless REA transformed the reply, which only the
file-offset mapping does; that case keeps Hopper's original reply.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
inspect_binary_layout and inspect_recorded_crash passed the decoded report
as both result and rawResult, so every Evidence record carried the whole
report twice (and an MCP response four times, counting its text copy).
The decoded report is the only representation, so raw_result is now null,
as for other single-representation operations.
The real pwntools verifiers asserted the duplicate; they now assert null.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
createAnalysisExecution defaulted an omitted rawResult to a second
complete copy of the result. Every caller that had no distinct provider
representation therefore doubled its Evidence: decode_interface_builder,
inspect_keyed_archive, inspect_asset_catalog, trace_dylib_resolution,
inspect_native_dispatch_metadata, observe_native_calls,
observe_native_ui/capture_native_ui_scenario (including screenshots), and
small Ghidra/IDA lifecycle records. Sibling operations of the same
providers already recorded null.
An omitted rawResult now records null. Normalized results are unchanged.
MCP responses repeat Evidence in text and structured content, so each
response and its share of the 10 MiB transport budget shrink by the same
half; e.g. trace_dylib_resolution on a real app bundle drops from
3,217,119 to 1,612,524 bytes per representation.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
compare-web-captures and compare-web-screenshots parsed their arguments as
inline JSON only, unlike every other JSON-input command. Screenshot
artifacts and scenario captures embed base64 PNG bytes, so a typical
viewport screenshot exceeds the command-line length limit (128 KiB per
argument on Linux, 1 MiB total on macOS) and could not be compared from
the CLI at all, while MCP accepts it. A file path was reported as
malformed JSON, and screenshot schema failures returned no issues.
Each argument now accepts inline JSON or a JSON file path through
parseCliJsonInput; file failures keep its path diagnostics, malformed
inline JSON keeps the field issue, and screenshot validation reports
projected schema issues. The tests share one CLI JSON process fixture.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
trace_web_module_imports and trace_web_source_location reported a missing
or non-file manifest, source map or import map as artifact_operation_failed
("Artifact could not be read or written" / "unsafe or conflicting internal
path"). These are caller selections: report ENOENT/ENOTDIR and non-regular
files as invalid input on the selected field, and EACCES/EPERM as a host
access denial. A missing manifest-derived source file stays an artifact fault.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Incur's `alias` map registers one-letter short flags and already accepts
kebab-case spellings of camelCase options. Twenty options registered their
kebab-case long name as an "alias", so help and completions advertised
flags such as `-case-sensitive` and `-relative-path` that the parser reads
as stacked short flags and rejects with "Unknown flag: -c".
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
list_architectures ran lipo on every admitted target. For a PE DLL, ELF
binary or plist, lipo's refusal surfaced as execution_failure with
"Retry once; if it continues, run rea doctor", which neither can fix.
Refuse non-Mach-O targets as capability_unavailable before running
lipo, as inspect_macho already does.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The streamed JSON encoder delegated every token through one recursive
generator per nesting level and yielded each token separately. Writing
analyze-javascript-application output for Typora (873 MB) spent 22 of
66 s there. Encode with an explicit stack into bounded 16 KiB parts:
identical bytes, 3-6x faster encoding, and the same laziness, error
and surrogate-pair guarantees.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
After #1301 merged, main's duplicated-line share rose to 2.503%, over
jscpd's 2.5% threshold, so Static checks fail on every PR. The new OMP
test repeated the dialect test's imports and npx registration command.
Export that command from the shared client test fixture instead.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* feat(setup): register OMP during setup
Add OMP (oh-my-pi) as a setup, doctor and uninstall client. Setup writes
a type: stdio entry to the user-level agent mcp.json that OMP reads,
following PI_CONFIG_DIR, an absolute PI_CODING_AGENT_DIR and the
OMP_PROFILE/PI_PROFILE named profile. It removes rea from OMP's
disabledServers denylist, which would otherwise hide the registration,
and doctor reports a listed registration as stale.
Closes#1257
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(setup): honor OMP's enabledServers allowlist in doctor
OMP's user enabledServers allowlist forces an entry with enabled: false
on, unless disabledServers also lists it. Doctor reported such a working
registration as stale. Also format the FAQ client list.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(setup): keep OMP's allowlisted entry in update maintenance
rea update skipped an owned OMP entry with enabled: false even when
OMP's enabledServers allowlist runs it. Apply the same forced-enable rule
as doctor. Share the client-location environment reset across the client
tests instead of repeating it in each file.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
trace_dylib_resolution roots and capture_process_scenario environment
names still advertised negative look-ahead patterns. RE2 and Rust schema
validators reject look-around, so a client using one refuses the whole
tools request, as #1282 reported for the NUL escape. Spell both
constraints as equivalent look-around-free patterns and assert that no
advertised pattern uses look-around or backreferences.
Closes#1282
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
