* feat(pe): inspect import and delay-load tables
Add inspect_pe_imports / inspect-pe-imports to read PE32/PE32+ import
and delay-load directories without executing the image. The report keeps
DLL names, named and ordinal functions, ILT/IAT RVAs, and a pefile
import hash of the standard import table. A missing directory is
observed absence, not proof the image never imported APIs.
* fix(pe): default PE import fixture options
Allow peImportImage() without an argument so CLI and filesystem tests
can use the KERNEL32 sample without repeating an empty options object.
* test(pe): keep import path cases on the dedicated MCP contract
inspect_pe_imports already rejects relative MCP paths in its own
boundary test. Leave the shared absolute-path matrix to tools whose
handler issue projection matches that suite's expected shape.
* fix(native): index virtual-to-file mapping lookups
mappedFileOffset scans every mapping twice per span, so a parser that
translates one span per table entry costs sections x entries. Add
fileMappingLookup, which builds an O(n log n) interval index once and
answers each span in O(log n) with the same accept, reject and
translation results. A differential test checks it against
mappedFileOffset over randomized overlapping mappings and zero fills.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ra7LM2NyQ7DumdssxbUq5g
* fix(pe): report partial import evidence and exact pefile hashes
- Withhold imphash with an explicit reason whenever the standard import
table is partial (unreadable names or thunks, missing tables), or when
pefile 2024.8.26 would hash a different row sequence: invalid or
truncated names, ordinal 0, descriptors without functions, tables its
bounds or repeat/spread heuristics discard, and its 8192-thunk import
symbol cap. Delay-load issues do not withhold it.
- Hash ordinal imports with pefile's complete ordlookup tables and its
ord<decimal> fallback instead of declining unmapped ordinals.
- Report a nonterminal descriptor whose FirstThunk is zero, and a
delay-load descriptor whose IAT or INT RVA is zero, as malformed_entry
partial coverage instead of a complete empty inventory.
- Directory evidence lists the file spans actually consumed for the
descriptor array and terminator; discontiguous mappings produce
separate spans and separate Evidence ranges.
- Carry exhausted limits as PeImportLimitFailure.limit and derive
recovery from it; the fixed 16 MiB retained-text ceiling reports a
fixed allowance instead of max_file_bytes advice.
- Translate RVAs through the indexed mapping lookup and yield with a
cancellation check every 1024 table entries.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ra7LM2NyQ7DumdssxbUq5g
* fix(pe): read only own ordinal tables for DLL names
DLL names such as constructor and __proto__ pass pefile's name rule but
resolved inherited object members, throwing before the ord<n> fallback.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ra7LM2NyQ7DumdssxbUq5g
* fix(native): preserve PE import facts and prove hash compatibility
* test(native): verify PE import workflows and measured costs
* refactor(native): narrow PE facts across the container union
* docs(native): record integrated PE import validation
* fix(native): bound every PE import RVA span
* docs(native): measure RVA guard completed workflows
* fix(pe): reconcile import coverage with retained rows
* fix(pe): retain partial imports at descriptor mapping gaps
---------
Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
* feat(ghidra): bound one procedure's decompile without closing the session
Fixes#5340
* fix(ghidra): report decompile limit as a request timeout
Keeps BinarySession from marking every Ghidra operation unavailable after one procedure's decompile reaches the limit, and names the MCP tool-call deadline in the remediation.
* fix(ghidra): budget the decompile limit and report cancellation-closed sessions
Adds REA_GHIDRA_DECOMPILE_TIMEOUT_SECONDS to the derived Codex/Kimi tool deadline and forwards it through setup. Marks a remote failure that settles after caller cancellation so the timeout projection no longer promises session reuse.
* feat(ghidra)!: default analyze_function to a retained summary
analyze_function was the one analysis tool never wired into REA's
detail: summary|complete convention, so it always returned the complete
native dossier inline (0.2-10 MB, dominated by value-flow and
references). Agents on tool-result-capped harnesses never reached the
pseudocode.
Add `detail` to analyze_function, defaulting to summary: return a
compact dossier (procedure identity, pseudocode size, facet counts,
value-flow availability) and retain the complete dossier Evidence in
the session. Page pseudocode, references, assembly and value-flow
through normalized_result.parent_evidence_id with inspect_analysis_view,
which already projects every native facet. detail: complete returns the
full dossier inline as before. Delivery detail never reaches the
provider or Evidence identity.
BREAKING CHANGE: analyze_function now defaults to the compact summary
view instead of the complete dossier inline. Pass detail: "complete"
for the previous behavior.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* feat(cli): expose analyze_function detail on the function command
Wire `rea function` into the shared CLI detail-delivery convention
(--detail, --evidence_output) so the CLI and MCP adapters offer the same
summary | complete choice for analyze_function. Summary delivery writes
the complete dossier Evidence to a bundle and returns the compact
summary view; it fails before analysis when --evidence_output is
missing, so no provider work is wasted. The CLI default stays complete
because a one-shot command has no in-process Evidence retention.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* refactor(ghidra): harden analyze_function summary after independent review
FlatlineRoundtable cross-model review (6 lanes) converged on the summary
projection's availability semantics being underspecified. Document that
facet_availability is authoritative over raw pseudocode length, that
code_units are UTF-16 (the unit inspect_analysis_view pages by), and that
value_flow.truncated is meaningful only when value-flow is available.
Parenthesize the nullish-coalescing/comparison for clarity. No behavior
change; declined claims (value-flow "truncated collapse", unresolved_calls
dead code) were schema-impossible or already correct.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* test(ghidra): cover analyze_function detail wiring on both adapters
Close the roundtable's top convergent finding: the summary path had only
domain-projection coverage, not the server/CLI wiring where the breaking
behavior lives.
MCP server (registerEnhancedTools): the default delivers a retained
summary whose complete parent is recorded for paging; delivery detail is
stripped before the provider request; detail: complete returns the full
dossier inline; a registration without retention rejects a summary
request before any provider work.
CLI (rea function): the default returns the complete dossier; --detail
summary without --evidence_output fails before analysis; with the bundle
sink it returns the compact summary and writes the complete Evidence
bundle.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(ghidra): keep analyze_function delivery detail out of provider inputs
Address the five automated review findings on #5660.
P1: `detail` lived in the shared enhancedInputSchemas.analyze_function, so
inspect_native_api composed `{procedure, detail: "summary"}` and the strict
Ghidra analyze_function schema rejected it. Delivery detail is now owned by
the analyze_function MCP tool contract; the handler separates it before any
provider call, and executeFunctionAnalysisEvidence parses the provider input
only. The MCP regression test validates provider input exactly as the
production Ghidra client does and exercises inspect_native_api.
P2: the analyze_function summary Evidence now carries the parent's
relationship_coverage like native facet views, so partial or unknown
caller/callee discovery is not reported as complete.
P2: the summary reports the dossier's declared facet_availability, nulls the
counts of facets the provider declared unavailable, and lists each reason in
unknowns, so Kuna's empty references, comments and basic blocks are not
mistaken for observed absence.
P2: inspect-analysis-view evidence-file sources accept the Evidence bundle
that --evidence_output writes, selecting a record with source.evidence_id
when the bundle holds several; the CLI test reads the saved bundle back
through that route.
Tests that assert the complete dossier over public analyze_function now pass
detail complete, and the oversized-dossier test asserts the summary view it
previously expected to be unsupported.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(ghidra): keep omitted function facets unknown in the summary
The analyze_function summary coerced an omitted unresolved_calls record to
a count of 0 while the native facet view reports the same dossier as "not
recorded", and it carried no counts at all for the pageable
swift_string_literals facets. Count optional records as null when the
dossier omits them, name them in unknowns, and add nullable counts for both
Swift literal facets. The summary Evidence envelope also dropped the
function address: createAnalysisViewEvidence now derives one function-view
predicate for both the address location and relationship coverage.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(scripts): request complete analyze_function dossiers in real-provider checks
The real-provider check scripts call analyze_function over MCP and then
consume the complete dossier: requireFunctionDossier, functionDossierSchema,
.procedure/.pseudocode reads, operation assertions, compare_functions inputs
and the oversized-dossier rejection path. The summary default broke them.
Every such call site now passes detail: "complete". Provider-direct callers
are unchanged because detail is an MCP contract field; the admission-error
sites stay unchanged because provider failures return before the summary
branch.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(ghidra): keep unavailable value-flow counts unknown in the summary
summarizeDossier coerced the five value_flow_* counts to zero when
native_value_flow was null (IDA, Kuna) or declared unavailable (Ghidra),
so the summary reported empty collections while value_flow.available said
otherwise. The counts are now null unless value flow is available, with the
reason already named in unknowns.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(evidence): budget authentication headroom by file shape and serialized size
The evidence-file headroom reserved two parsed copies, calibrated for one
record, and charged strings at two bytes per character. Bundle
authentication additionally holds the schema-parsed bundle, the re-parsed
records and two JSON.stringify results, which expand control characters and
lone surrogates sixfold: a bundle whose raw_result is a million NUL
characters estimated at 2 MB parsed but needed 12 MB of serialized text.
The budget now has a parsed term and a serialized term per file shape, owned
by EvidenceAuthenticationHeadroom and shared by the analysis-view reader and
the bundle importer, which had no guard. Two transient peaks the budget did
not name are removed at their producers: the canonical digest hashes a large
leaf directly instead of concatenating it into its buffer, and the JSON
validator walks containers by cursor so a wide array no longer queues one
entry per member.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(mcp): route analyze_function summary replies to dossier consumers in docs and prompts
The MCP guide, the version-comparison and crash-triage prompt contracts,
the authored skill and the Kuna, IDA and function-matching docs still told
agents to pass analyze_function replies straight to compare_functions or
build_call_path. Under the summary default those consumers reject the reply.
They now name the retained-evidence reference for compare_functions, the
observations fields of match_native_functions, and detail: "complete" for
build_call_path, which accepts only inline dossiers.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* test(evidence): pin validator member order and keep interrupted validation yielding
Independent review of the cursor validator found the interrupted-validation
fixture relied on the old last-first walk: its over-deep head was reached
before the first cooperative yield. The padding now precedes the deep
element so the walk yields first, which is the invariant the test guards.
A new case pins first-defect-in-member-order reporting, skipped
non-enumerable keys no longer count toward the yield cadence, and the
import-headroom test loads the reader through a fresh module graph so it
holds under the services project's shared workers.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(ghidra): align deadline bounds and cancellation outcomes
* refactor(ghidra): shorten deadline recovery guidance
* fix(mcp): clarify the scope of inferred decompile deadlines
* test(ghidra): verify deadline workflows and document measured costs
Separate failure and registration cases by their distinct contracts.
Record actual Linux Ghidra, CLI and MCP evidence with measured context,
elapsed time, CPU and memory, and explicit remaining coverage limits.
* docs(ghidra): pin integration and distinguish baseline format failures
* fix(ghidra): own failure lifecycle wording in the provider
* test(ghidra): budget timeout fixture generation before allocation
* fix(analysis): preserve summary provenance and bound bundle guidance
* fix(ghidra): align capability and traversal deadline guidance
* style(config): align deadline guidance formatting
* docs(ghidra): record capability and traversal verification
* docs(analysis): use absolute paths for Evidence files
* docs(analysis): record completed function summary measurements
* docs(analysis): update main integration validation
---------
Co-authored-by: Syed Rishad <rishadzzzz@gmail.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
Resolve application and managed workflow contracts through the canonical name-based resolver, removing catalog-position overloads and the obsolete local facade.
Use the existing retained-summary policy for eight offline registrations. Pass only validated delivery detail to that owner while preserving exact typed service inputs and validation, admission, logging, and Evidence recording order.
Align export-shape MCP parity coverage with the existing absolute-path contract using an absolute dot alias. Preserve exact caller-path parameters, canonical artifact parity, distinct Evidence identity, and the export comparison assertions.
Route MCP snapshot reads through the session's existing file-resource owner so cleanup remains visible, blocks replacement, and can be retried by close_binary. Keep reads inside the serialized open callback and remove superseded adapter resource plumbing.
Validate typed errors, exact descriptor recovery, unchanged source bytes, replay provenance, and same-connection lifecycle ordering. Focused snapshot, target cleanup, and MCP lifecycle checks passed with simulated providers; no live provider was exercised.
Fixes#6216
* feat(decompile): write function pseudocode to a directory and return a manifest
Add decompile_all and `rea decompile-all` so a caller can write one
provider pseudocode file per function, or a filtered set, and receive
only a small manifest. The result stays bounded when the program is
larger than a local model's tool-result cap.
Resume continues an export for the same target digest, provider,
analysis profile, and layout. Matching files are skipped, conflicting
files stay unchanged, and one function's failure does not stop the
rest. Cancellation leaves files already written. Requesting dwarf
layout reports that compilation-unit paths are unavailable.
Simulated-provider tests cover the manifest, resume, filters, limits,
cancellation, and the failure cap. MCP registration, CLI coverage,
schema validity, the product catalog, check:fast, and docs:check
passed. Real Ghidra was not exercised.
Refs #6090
* fix(decompile): keep unknown export identity unresumable and recover torn journals
Resume treated two missing target or profile digests as a match, so an
export of one binary could continue into another. Resume now requires both
digests, and a decompiler observation without the export's profile, or with
a different target, is recorded as profile_changed.
A crash during a journal append left a torn final line that made the export
unresumable. Text after the last newline is now treated as uncommitted and
truncated before the next append, and stale temporary files are removed.
An inventory row with an empty address aborted the whole export because its
failure record failed the journal schema. Such rows are now excluded and
counted in the limitations.
A limited resume retried the same recorded failures first and never reached
new functions. Functions not yet attempted now run before retries. File path
reservations come from every journal line, so an output_io record no longer
releases a path. The index is rebuilt from the journal in numeric address
order, and journal close and cleanup failures return a Result.
The cancellation test now cancels after a file is committed and checks that
file survives and is skipped on resume. Redundant snapshot-replay checks are
removed, the CLI limit uses the contract constant, and --snapshot no longer
claims decompile-all updates it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(decompile): refuse resume of a missing export and repair journals by byte offset
A second review round found that resume against a directory that does not
exist silently started a full new export and reported it as resumed. Resume
now requires an existing export directory.
The torn-line repair measured the committed journal from decoded text, so
damaged UTF-8 could move the truncation point past the end of the file. The
journal is now read as bytes and cut at the last newline byte, and an intact
journal is not truncated, so a concurrent append is kept.
A per-function observation for a different target is now recorded as
target_changed instead of profile_changed. A failure to remove a partly
created output directory is reported instead of hidden. The journal handle is
closed on every path again, including an unexpected throw. The path allocator
no longer gives up after 1,000 collisions and reuses a taken name.
Tests now use createTestTempDirectory, assert journal lines against the
journal schema, and cover resume of a missing directory, resume with an
unknown analysis profile, and a target-mismatched observation.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(decompile): keep a function's first file name across failed retries
Reserving every journal path stopped other addresses from taking a path, but
an address looked up only its latest record. After its file went missing and
a retry failed without a path, the next resume moved the function to a
hash-suffixed name and orphaned the original. Each address now keeps the
first path the journal gave it.
A matched commit still ran the decompiler but did not count toward limit.
The limit now counts every function taken on after skipping verified files.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(decompile): never follow a journal symlink and check procedure_info identity
Torn-line repair truncates the journal, and the journal was opened through
any symlink in the export directory, so repair could truncate a file outside
it. The journal must now be a regular file and is opened without following
symlinks where the platform supports it.
Body ranges from procedure_info were written without checking that the
observation belonged to this export or described the requested function. An
observation with a different provider, profile or target, or another entry
point, now leaves the body unavailable with that reason.
File, journal and index writes now use FileHandle.writeFile, which writes the
whole buffer instead of trusting one write call. The stale temporary sweep is
removed: a matching file name does not prove this export created the file.
The inverted-range test now uses one address space so it reaches the
inversion check.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(decompile): bound returned caveats and keep verified checksums
Ghidra names each terminal call's address in its per-function caveats,
so decompile_all copied one new caveat per function into the returned
Evidence and the answer grew with the program. The returned Evidence
now keeps the eight most often observed distinct per-function caveats,
grouped by provider and text as they are rendered, plus a count of the
rest. Inventory caveats are unchanged, and other workflows are not
bounded. Each function keeps the caveats of the observations it used:
a written file in its header, now including procedure_info caveats,
and a failed attempt in its journal and index record. A procedure_info
observation for another target or profile is foreign and contributes
none.
A failure record for a function that was already written replaced the
record holding its checksum, so after a transient read error a resume
decompiled that function again. The journal now keeps the last
successful write separately from the latest attempt. Resume verifies
against that write and records it again once the file verifies.
A caller abort that arrives while the export finishes now returns
cancellation with the caller's reason instead of a success manifest,
as #6143 does for other late finalization. A publish, index or close
failure stays the primary error.
Also make the missing-journal resume error say what to do, and add
boundary tests: one runs decompile-all through the real CLI against a
fake Kuna to check flag mapping, the relative output directory,
filters and limit; one runs decompile_all through production MCP
registration to check that the caveat bound is applied.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(decompile): keep each function's complete caveats in the index
The returned Evidence keeps eight distinct per-function caveats and
counts the rest, pointing readers at the export. A written function's
file header truncates its caveats after 4,000 characters, and its
journal record kept none, so a large function with many terminal calls
could lose caveats from both the answer and the export.
Every journal record, written or failed, now keeps the complete caveats
of the observations its attempt used, and decompile-index.json carries
them. The header stays a readable summary and, when truncated, says
that the index keeps the complete list.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(decompile): report external functions instead of failing them
Ghidra classifies imports as external functions with an empty body and
returns null pseudocode for them, as its limitations state. decompile_all
recorded each one as an invalid_pseudocode failure ("not text"), so a
dynamically linked program reported every import as a decompiler failure:
34 of 116 functions in /usr/bin/xxd.
procedure_info owns the external classification. When it classifies a
function as external, the export now records status "external", counts it
in counts.external, writes no file and requests no pseudocode. The record
keeps the body fact as observed. The journal header pins the target,
provider and profile, so a resume skips a recorded external function
without asking the provider again; a limit below the external count still
leaves nothing remaining. A null pseudocode result for a function whose
classification is not external is reported as no_pseudocode.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(decompile): report finalization failures after a cancelled export
When the caller's abort stopped the publish loop, decompile_all returned
that cancellation before checking the index write and the journal close,
which had already run. A failure to finalize the export was hidden, so
the caller could not tell the saved index or journal was incomplete.
Publish failures stay primary, then index and close failures. A
cancellation counts as the caller's only when the caller's signal is
aborted, and is then reported after a successful finalization with the
caller's own reason. A provider's own cancellation, without a caller
abort, stays a publish failure like any other. Tests cover an index
failure and a close failure after the caller stops the loop, a stopped
loop with a successful finalization keeping the caller's reason, and a
provider cancellation staying primary over an index failure.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(decompile): remove a created output directory that cannot be secured
createPrivateDirectory ran mkdir and chmod as one step. When mkdir
succeeded and chmod failed, it returned an error before #create reached
its cleanup, leaving an empty directory behind. A fresh export then
refused the existing directory, and a resume refused it for having no
journal, so the caller had to remove it by hand.
mkdir and chmod are now separate. A mkdir failure created nothing, so
nothing is removed. A chmod failure after this call created the
directory removes it with a non-recursive rmdir: it is still empty, so a
path another process filled or replaced meanwhile is refused rather than
deleted, and that refusal is reported.
A filesystem boundary test fails chmod for the new root and checks the
directory is gone and a retry succeeds. A module test confirms that a
function written earlier and now classified as external still reports a
changed file at its path as output_exists.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* refactor(process): share the socket lease with owners other than Hopper
The socket lease that keeps two REA processes from opening the same Hopper
document lived only in HopperLease, with Hopper error messages built in.
A decompile export needs the same guarantee for its output directory.
OwnedLease now holds the lease mechanism unchanged: a listening local
socket that the operating system ends with its process, owner identity
answered by the live holder, and recovery of a stale socket serialized
through a recovery directory. Callers map a tagged LeaseFailure to their
own error. HopperLease keeps its exports and maps each failure to the
same HopperStartError message as before. On Windows the lease is a named
pipe, which needs no stale recovery.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* fix(decompile): hold the output directory per run and stop on repeated write failures
Two runs could resume the same output directory at once. Each read the
journal once, and the index each wrote reflected only its own view, so
the last writer could drop the other run's records. A run's torn-line
repair could also truncate the other run's append.
DecompileExport now holds a lease on the output directory, named by its
device and inode, from before the journal is read (or, for a new export,
from just after the directory is created and before its journal exists)
until the journal is closed. A second run is refused as a retryable
capacity constraint that reports the holder's process ID. The lease ends
with its process, so a crashed run leaves no lock to clear. The lease is
released on every open failure, and a failed release is reported.
An export also kept decompiling every remaining function when its files
had stopped being writable mid-run, for example on a full disk, failing
each one with output_io. After five consecutive functions whose files
could not be written or read, the export now stops with an output error
that carries the counts and the last failures. A verified file or any
other outcome ends a run of failures. Committed files and records stay,
the index is still written, and a later resume continues.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AoJZDPcrgmMsCqLG4LfiFb
* docs(native): record directory export verification
Record matched real Ghidra outputs, bounded caveat and cleanup regressions,
actual catalog-inclusive context estimates, runtime measurements and coverage
limits for the pinned integration.
* fix(native): bound export summaries and reserve portable paths
Use one case-folded reservation owner for allocation and resumed journals.
Budget inline failures and provider caveats by encoded JSON bytes while
retaining complete records in the output index. Align the tool contract
with the collector order and verify SDK result-window behavior.
* docs(native): refresh final export measurements
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
Artifact comparison owns the inventory Evidence identities whose integrity mismatched. Residual unknowns consume those identities and static/runtime correlation keeps contradictory or unresolved findings separate from observed changes.
Adds a real-ASAR MCP comparison and correlation regression with ordinary-content controls. Validation is pending the centrally coordinated lane; single-file formatting and lint completed.
Reuse validated class pages for inline pagination audits while retaining complete provider replies. Move retention into the shared Android workflow and remove the MCP-only service wrapper and application text reparser. Preserve unknown JSON metadata, late cancellation and retention failures.
Route CLI and MCP coverage requests through the shared strict workflow boundary before authenticating source Evidence. Preserve canonical advertised schemas, use integrity errors for malformed source bundles, and use the workflow operation for input diagnostics.
Remove the unused throwing domain facade and keep semantic fixture assertions on the parsed Result APIs. Validate compiled CLI and stdio MCP parity, strict request rejection, domain authority/time semantics, generated schemas, and unexpected failure propagation. No external providers execute in these checks.
Separate source corruption and selected-role errors from unexpected derivation failures. Normalize coordinates before body membership and conflict checks, reuse parsed requests, and account for outer observation arrays before validation copies.
* fix(mcp): report Linux-only integrations as unsupported_host off their hosts
Session discovery folded the host check into each optional integration's
enabled flag, so on macOS or Windows the Linux-only packet, signal,
firmware, pwntools, EVM and Wakaru tools reported provider_missing.
Their execution guards refuse those hosts whatever is configured, so
installing a provider there could never make them available.
The server now records, per integration, when the host is outside its
verified host and no service is injected. Discovery checks that before
configuration and reports unsupported_host with the existing
remediation. Injected services and supported hosts behave as before.
Fixes#5730
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* refactor(mcp): share provider host checks with session discovery
Discovery restated each Linux-only integration's host rule in
createServer while its provider kept a separate execution check, so the
two could drift apart again. Each provider now exports its host
predicate, uses it in its execution guard, and composition hands the
same predicate to createServer for both the host restriction and the
enabled flag.
Observed host restrictions now also apply when an embedder supplies its
own availability policy, as provider load failures already do, because
no policy lets the default providers pass their host checks.
The MCP regression moves to tests/process-global and emulates darwin
arm64, linux arm64 and linux x64 through the real createServer, so CI on
Linux x64 exercises the unsupported-host branch. It also covers a
configured policy and the matching execution refusals.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(mcp): keep caller host restrictions beside observed ones
The observed host restriction replaced a configured policy's whole
hostUnsupported map. For an injected service the observed value is
false, so a caller's own restriction was lost and the tool was reported
available. Combine both per integration: either the caller or the
observed host can mark it unsupported.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(pcap): expect unsupported_host for packet tools off Linux
The host-independent PATH case asserted provider_missing on every host.
Off Linux the packet tools are unavailable because of the host, so
discovery now reports unsupported_host there; Linux still reports
provider_missing for an unusable PATH.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Resolve raw input at the application boundary and reuse its facts in both MCP admission paths. Keep cleanup, mutation retirement and execution ordering while retaining structured correction details for rejected commands.
Validation: 186 focused tests, source checks, docs conformance and duplication gate passed. The new compiled CLI regression fails on the main baseline. Provider process controls use deterministic fixtures; no real Rizin verification is claimed.
Parse caller selections and authenticated Evidence at their owners, reuse the parsed function and comparison facts, and preserve unexpected implementation failures. Remove downstream error guessing and redundant validation wrappers.
Verify with 526 affected tests, source checks, docs conformance, test discovery and duplication checks. Seven public SDK regressions fail on the main baseline. No real-provider execution was performed.
* refactor(pcap): give TShark selection and version admission one owner
The pcap and packet-capture families wrapped the same caller-supplied
TShark with diverging rules. inspect_pcap_* accepted an absolute
REA_TSHARK_COMMAND or tshark from PATH, while inspect_packet_* and the
server availability policy required an absolute REA_TSHARK_COMMAND and
parsed the version banner with a separate regex.
PcapConfiguration now owns the rule: selectWiresharkTool selects an
absolute environment path or the bare name on PATH and refuses relative
values; locateWiresharkTool resolves a PATH selection to the first
absolute PATH entry holding an executable file; parseWiresharkVersion
admits Wireshark 4.x banners for both providers. TsharkCaptureProvider
and the session availability policy consume these facts instead of
reinterpreting REA_TSHARK_COMMAND, so all four tools accept and refuse
the same configurations. inspect_packet_* still resolves, hashes and
rechecks the located executable, and records selection_source in its
tool Evidence.
Host support and resource bounds are unchanged; unifying them is the
next milestone in #3403.
Refs #3403
* fix(pcap): execute the TShark the shared locator selects
resolvePcapTools still returned bare tool names for PATH selections, so
the operating system could run a tool from a relative PATH entry while
the packet tools skipped that entry and selected a different install.
On POSIX hosts, resolve PATH selections through locateWiresharkTool and
run that absolute executable, failing with the existing tshark_missing or
capinfos_missing code when no absolute PATH entry holds the tool. Both
selections are validated before PATH is consulted, so a relative
environment value still reports its own configuration error. Windows
keeps PATH/PATHEXT resolution in the process launcher.
Refs #3403
* chore: restore main's copies of files the merge hook reformatted
The pre-commit oxlint --fix touched two files this PR does not change; keep them identical to main.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(pcap): share selected tool identity verification
* perf(pcap): retain compact executable identity
* fix(pcap): report missing selected tools clearly
* fix(pcap): read historical capture evidence
* fix(pcap): preserve executable read failures
* test(pcap): cover version refusal and packet-tool availability selection
Address review follow-ups on the shared TShark owner:
- Send a 3.x TShark banner and a non-Wireshark banner through the packet
provider, pinning the capability-unavailable reason for each and that
only the version probe ran before refusal.
- Cover the session availability rule for inspect_packet_capture over
MCP: a PATH-only install and an absolute REA_TSHARK_COMMAND are
available on Linux, a relative REA_TSHARK_COMMAND is refused even when
PATH holds TShark, and no absolute PATH entry reports provider_missing.
- Rename the locator test's missing absolute PATH entry so it no longer
shadows node:path relative.
Refs #3403
* fix(pcap): name the configured path when a selected tool is missing
Resolving the selected executable before launch replaced the runner's
"Cannot run selected tool <path>" refusal with a message that kept the
configured path only in diagnostics, so the public CLI no longer told the
caller which REA_TSHARK_COMMAND or REA_CAPINFOS_COMMAND value to fix and
the pcap CLI boundary test failed. Include the configured command in the
capability-unavailable message and assert it there.
Refs #3403
* test(captures): give the late-cancellation fixture a TShark selection source
The service fixture added on main in 719d56ebf predates the required tool.selection_source field, so its report failed output validation before reaching the late-cancellation path the test covers.
Refs #3403
* fix(pcap): report a probe's pending cleanup when a post-run check fails
When a tool probe returned with its owned process still awaiting cleanup,
a failed post-run executable re-read or identity check threw before run()
reported that cleanup. Probes have no snapshot to carry it, so callers
saw an ordinary error with cleanupIncomplete false while the provider
kept the process and refused the next request.
After the command returns, combine such a failure with the probe's
ProviderCleanupError (cancellation and other command failures keep their
own cleanup outcome). Classify request failures through one
analysisFailure owner shared by run() and executeAdmitted(), and let the
availability probe rethrow any cleanup-incomplete failure instead of
reporting the tool unavailable.
Refs #3403
* fix(pcap): keep a failed executable re-read's cleanup live with the probe's
The combined post-run failure was bound only to the probe's pending
process cleanup. When the executable re-read had also failed to close its
file, the classified failure owned that cleanup, but the combined error's
live cleanupResources omitted the file and could clear cleanupIncomplete
once the process stopped. Bind both the classified failure and the probe
cleanup as dependencies.
Refs #3403
---------
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
* fix(wasm): refuse unretained summary delivery before provider work
inspect_wasm_artifact with detail summary ran the WASM provider before
checking whether the server retains Evidence. On a fixed server without
session retention a successful provider run was followed by
capability_unavailable, and a failing provider run replaced that
diagnostic with the provider error, hiding the guidance to select
detail complete.
Check retention right after input validation, as the EVM, pcap and
network-capture handlers do, so the refusal precedes any provider work
and keeps its typed diagnostic. Complete delivery is unchanged. The
fixed-server boundary test now injects a vi.fn provider and asserts it
stays idle for refused summaries while complete delivery still runs it.
Fixes#5100
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(wasm): validate caller paths before the summary retention preflight
Review follow-up for #5274. The retention preflight ran ahead of the
semantic input checks, so a relative path, glue path or candidate path
with detail summary on a fixed server returned capability_unavailable
instead of its field-specific invalid_request.
WasmArtifactService now exposes that parse-and-absolute-path step as
validate(); inspect() uses the same method, so the checks stay at the
owning boundary, and the MCP handler validates before deciding on
retention. The composition descriptor and server typing expose validate.
The fixed-server boundary test sends relative path and glue_paths with
detail summary and asserts invalid_request with both issue paths and no
provider call; it failed on 418a66d7d with capability_unavailable.
docs/mcp-contracts.md now claims the before-analysis refusal only for
inspect_artifact, inspect_wasm_artifact, inspect_macho and
decode_interface_builder, which enforce it; inspect_binary_layout and
analyze_javascript_application still analyze first and are left to a
separate change.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* refactor(wasm): inspect validated artifact input without reparsing it
Follow-up for #5274, matching the review outcome on #5288. The MCP
handler validated the request and then called inspect(), which ran the
same validator again, so every call parsed and cloned its input twice.
WasmArtifactService now separates the raw entry point from execution:
inspect() validates raw input once and delegates to the new
inspectValidated(), which runs the inspection for input that validate()
already admitted. The handler validates, runs the retention preflight,
then calls inspectValidated(); the composition descriptor and server
typing expose it. The CLI keeps the raw inspect() entry point; behavior
and diagnostics are unchanged.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(wasm): keep cancellation ahead of raw artifact input validation
Follow-up for #5274, matching the review outcome on #5288. Separating
validate() from inspectValidated() moved the raw inspect() entry point's
validation ahead of its abort check, so a direct call with an
already-aborted signal and malformed input returned AnalysisInputError
instead of the AnalysisCancelledError the pre-change service gave.
The raw entry point now checks the signal before validating, as before,
while inspectValidated() keeps its own guard for the MCP handler path.
A unit test pins the precedence with no provider call.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
* fix(mcp): escalate transient failures before asking for a restart
Give dropped connections and startup failures one bounded retry, then a
scripted handoff to the user. Point timeout recovery at a named request
field or REA_* setting instead of telling agents those limits are fixed.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(process): stop unexpected command failures from recommending a retry
The catch can already have launched the target, and the projection stays
non-retryable. Tell the user to check the inputs and run doctor instead.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(mcp): limit retries to safe calls and recommended settings
A dropped connection may already have changed the target, and a bad
catalog rejects the same startup every time. Retry only a repeatable
call, and change a setting only when the remediation recommends it.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* test(mcp): match recovery guidance across wrapped lines
Startup handoff capitalization and markdown wrapping still carry the
same sentences. Compare the prose with whitespace collapsed.
Co-authored-by: morluto <morluto@users.noreply.github.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
* feat(python): inspect CPython pyc headers and marshal code objects
Add inspect_python_pyc / inspect-python-pyc to read one explicit local
CPython 3.7–3.13 .pyc without executing bytecode or recovering source.
* chore(python): drop unused PythonPycHeader export
Knip reported the type as unused after inspect_python_pyc landed.
* fix(python): bound pyc expansion and decode marshal like CPython
- Count a nested code object reached through TYPE_REF once per copy the
report carries, against max_objects and max_file_bytes, before counting
or validating the tree
- Charge long digits to max_string_bytes, reject out-of-range and
unnormalized digits, build the value in one pass, and report integers
above 256 bits by bit length only
- Decode TYPE_UNICODE with surrogatepass semantics and reject other
ill-formed UTF-8
- Map only 3.11 magics 3494-3495 to the py311 layout
- Report non-finite float and complex parts as inf, -inf or nan
- Allow an empty co_filename
- Require bytes for co_linetable and co_exceptiontable
- Bind cleanup-failure errors to their retained descriptor leases
* fix(python): list pyc code objects once instead of a recursive tree
Report each distinct code object once in code_objects, top-level first,
with nested_code as indexes. A child shared through TYPE_REF is no longer
copied per reference, and the MCP output schema is no longer recursive.
Names repeated through TYPE_REF still count against max_file_bytes for
each copy the report carries.
* fix(python): bound shared pyc constants and reject malformed tables
- Cache byte-string digests so a buffer shared through TYPE_REF is
hashed once instead of on every reference.
- Charge every reported co_consts entry, including tables shared
through TYPE_REF, against max_objects before summaries are built.
- Resolve the inspect-python-pyc path with absoluteCallerPath so
link/../file.pyc reads the file the filesystem selects.
- Reject TYPE_NULL inside marshal sequences, as CPython does.
- Reject a co_localspluskinds length that does not match
co_localsplusnames in the parser and the report schema.
* fix(python): bound shared pyc names and read CPython marshal graphs
Charge each reported code-constant name, keep integer bit lengths on the
decoded value, and list code objects nested inside constant containers.
Accept self-referential lists and dicts, marshal version 0 text floats,
empty code names, the original string byte count, and Python set and dict
lengths. Pass the CLI cancellation signal into command cleanup.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(python): bound pyc key expansion and shared nested code
Charge set and dictionary keys before a shared tuple doubles them, list each
nested code object once per owner, and accept code objects as CPython keys.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(python): correct pyc equality, counts, and nested walks
Summarize code objects only after the marshal graph is complete, and
collect nested code with an iterative walk that shares wrapper results
instead of copying every descendant list. Keep tuple boundaries and the
original string bytes in equality keys, and reject negative CPython counts.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(python): bound shared pyc names and keep the first dict key
Charge every copied name-table entry against the object budget before
building the report array, including empty strings that add no bytes.
When dictionary keys compare equal, keep the first key object and
replace only its value, matching CPython.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(python): walk each pyc dictionary child list once
Keep a container's child array on the nested-code frame so a wide
dictionary is not copied on every step of the synchronous marshal walk.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(python): reuse stored pyc dictionary children when combining
The nested walk keeps each container's child list on its frame.
Combining nested code rebuilt that list, and the 8,000-key regression
still accepted the old per-key copy. Use the stored list, and require
that dictionary to finish well under the copy's runtime while still
reporting a nested code object.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(python): keep distinct pyc code objects and reject corrupt ones
Encode each name table's length and entry boundaries in the equality
key so a different name list cannot share a memo key. Walk tuple and
frozenset keys on an explicit stack so a long TYPE_REF chain does not
overflow. Treat code objects CPython rejects for an odd co_code or an
undersized local table as malformed instead of complete coverage.
Co-authored-by: morluto <morluto@users.noreply.github.com>
---------
Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com>
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
When a file:// URI contains leading slashes representing a UNC path (such as file:////remote.example.test/share/state.json), Node's fileURLToPath on Windows throws ERR_INVALID_FILE_URL_PATH because it expects a drive letter or Windows UNC format. This was caught and misreported as 'a file:// uri names a local absolute path' instead of the intended 'a file:// uri names a path on this host, not a network share'.
Testing UNC_PATH against url.pathname before calling fileURLToPath ensures that UNC paths are consistently recognized and refused as network shares across platforms.
Co-authored-by: ithis0623-png <ithis0623-png@users.noreply.github.com>
* feat(erlang): inspect BEAM IFF chunks and tables
Replay the Erlang BEAM inspection workflow, including table bounds,
OTP 28+ atom layouts, the shared absolute-path policy, and empty-path
CLI coverage, without carrying GitHub workflow edits.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(erlang): use publicInputStrictObject for BEAM tool input
inspect_erlang_beam advertised a public input built with bare
z.strictObject, so later shared input-policy changes would miss it.
Result and internal objects stay on their own constructors.
Co-authored-by: morluto <morluto@users.noreply.github.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
* feat(pe): inspect Authenticode certificate tables
Add inspect_pe_certificates for IMAGE_DIRECTORY_ENTRY_SECURITY
WIN_CERTIFICATE entries and structurally decoded PKCS#7 signer
and X.509 identity fields. The workflow reads a stable PE image
without executing it or verifying signatures, chains, timestamps,
or Authenticode digest coverage.
* fix(pe): bound and validate certificate DER decoding
- Cap decoded DER values with a shared node budget so sibling fan-out
cannot exhaust memory before the PKCS#7 check
- Require SignedData version, encapContentInfo and a final signerInfos
SET in RFC 5652 order, and a signature OCTET STRING in each SignerInfo
- Decode OBJECT IDENTIFIER arcs as unbounded base-128 values, including
a multi-byte first combined arc
- Decode UniversalString names as UTF-32BE and reject invalid values
- Share PE header admission with the resource layout so a section table
that does not fit SizeOfHeaders is rejected
* docs: list PE certificates in the translated READMEs
* fix(pe): cap OBJECT IDENTIFIER length and arc size
A crafted type-2 certificate with a 262,144-byte contentType OID stalled
decodeOid for about 20 seconds, because one arc kept growing a BigInt
before the signedData check ran. decodeOid now rejects an OID longer than
256 bytes and an arc longer than 32 bytes before decoding it. Certificate
OIDs, including 2.25 UUID arcs, fit under both caps.
* fix(pe): reject truncated certificate tables and PKCS#7
SignedData now requires one SignerInfo with a non-empty signature and a
complete issuerAndSerialNumber. Certificate names share a 64 KiB decoded-byte
budget and are copied once. A security directory that ends in a short
WIN_CERTIFICATE header fails instead of reporting a complete table. The
certificate CLI forwards the caller abort signal into cleanup.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* chore(ci): keep fork workflow files unchanged
The certificate restack follows current main. Workflow files stay at the
fork tip so this branch update does not rewrite .github/workflows.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(pe): charge serial and validity text to the decoded budget
Certificate and signer serial numbers and validity times now count against
the shared decoded-text budget before they are converted, so a crafted
certificate cannot push megabytes of text into the hashed evidence. The
budget error now names certificate text rather than only names.
Restore the CI workflow to match main: the CI_TIMING_TIER input, the
tier-named timing artifacts and their publication conditions, the Windows
typecheck, and the path and snapshot suites in the curated Windows lane.
* fix(pe): require certificate fields and preserve signer key ids
Reject X.509 records that omit subjectPublicKeyInfo or signatureValue,
return subjectKeyIdentifier bytes as bounded hex, and advertise an absolute
path in the PE certificate input schema.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(pe): reject extra PKCS#7 wrappers and invalid UTF-8 names
ContentInfo and its signedData wrapper no longer ignore trailing values, and
directory strings must round-trip as UTF-8 instead of becoming U+FFFD.
Co-authored-by: morluto <morluto@users.noreply.github.com>
---------
Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com>
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
gensokyo-dev