* fix(hopper): diagnose MCP provider failures with stage, state, and request context
* fix(hopper): retain health at the actual provider lifecycle boundary
* test(hopper): distinguish socket loss from later observed process exit
* test(hopper): cover observed exit separately from socket-first replies
---------
Co-authored-by: N0zoM1z0 <1705004398@qq.com>
compareCodePoints documented a Unicode code-point comparison but used
</> operators, which compare UTF-16 code units. For astral code points
(above U+FFFF) the lead surrogate 0xD800-0xDBFF is numerically below the
BMP tail U+E000-U+FFFF, so those characters sorted before it.
Delegate to the existing compareUnicodeCodePoints, which the domain
already uses elsewhere, so the two canonical comparators agree. All call
sites use only the sign, so sort and '< 0' checks are unaffected.
Adds the missing regression coverage for canonicalOrdering.
Co-authored-by: TahsinArafat <TahsinArafat@users.noreply.github.com>
Co-authored-by: N0zoM1z0 <1705004398@qq.com>
* fix(browser): advertise zero-based units for runtime source coordinates
inspect_web_event_listeners and observe_web_execution return CDP
callsite, listener and resource_start line_number/column_number values
unchanged: zero-based lines and UTF-16 columns. Neither the advertised
output schema nor docs/web-runtime.md said so. trace_web_source_location
takes one-based lines, so a caller chaining a listener location into it
lands one line early.
Describe the units on the shared schema fields, which reach both tools'
advertised output schemas, and document the conversion.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(browser): describe runtime positions relative to the enclosing resource
For an inline <script>, CDP reports listener and callsite positions
relative to the HTML document, and resource_start is where the script
text starts in that document. A real headless Chromium 155 page showed
this: a listener on HTML line 6 came back as line_number 5, with
resource_start {3, 8}. Describe all positions as relative to the
enclosing resource. Document that inline-script positions must have
resource_start subtracted, plus one added to the line, before
trace_web_source_location.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(browser): clarify runtime-to-source coordinate conversion
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: N0zoM1z0 <1705004398@qq.com>
* test(javascript): reproduce HTML carriage return source coordinates
Compare actual resolved and unresolved artifact graph evidence with parse5 source locations across HTML line terminators and UTF-16 columns. Exercise both public built CLI analysis commands. Tests only; production and workflows unchanged.
Signed-off-by: Rudy Celekli <rudy@gradiahq.com>
* fix(javascript): preserve HTML carriage return source ranges
Treat LF, CRLF, and bare CR as HTML line endings when projecting actual script opening-tag offsets to retained source ranges. Preserve UTF-16 column semantics and document the source-coordinate contract.
Signed-off-by: Rudy Celekli <rudy@gradiahq.com>
---------
Signed-off-by: Rudy Celekli <rudy@gradiahq.com>
* test(javascript): reproduce BOM source range drift
Compare original UTF-8 source bytes and their retained digests with direct Babel and parse5 source-location oracles through actual directory and ASAR artifact reconstruction and both built public CLI analysis commands. Preserve no-BOM, structured-BOM and invalid UTF-8 controls. Tests only; production and workflows unchanged.
Signed-off-by: Rudy Celekli <rudy@gradiahq.com>
* fix(javascript): preserve original BOM source coordinates
Retain the initial UTF-8 BOM for JavaScript and HTML source text so graph ranges refer to the original digest-bound bytes. Preserve structured-text BOM handling and fatal UTF-8 decoding.
Signed-off-by: Rudy Celekli <rudy@gradiahq.com>
* style(javascript): format BOM decoder option
Apply the locked Oxfmt line wrapping to the approved decoder option; no behavior or test changes.
Signed-off-by: Rudy Celekli <rudy@gradiahq.com>
---------
Signed-off-by: Rudy Celekli <rudy@gradiahq.com>
* fix(process): name the reason two captures cannot be compared
compare-process-captures rejected captures whose comparison contracts
differ, but the reason never reached the caller. The CLI printed
"Process command could not complete... run `rea doctor`". MCP wrapped
the error in an AnalysisInputError with no issues ("Analysis input is
invalid. Check the arguments"). A stale capture under
max_capture_age_ms got the same treatment.
Throw typed input issues that name the differing contract fields, or
the stale side with its completion time and age, and let
runDerivedOperation pass AnalysisInputError through unchanged. Document
what the comparison contract covers.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(process): report CLI trace-spec issues under the MCP input path
The CLI reported an invalid trace specification under operation
"compare-process-captures" with root-relative issue paths such as
["events"]. MCP reports the same input as compare_process_captures with
paths under trace_spec. Use the canonical operation name and prefix
trace_spec so both surfaces describe the same input identically.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(process): name the side and constraint for unusable capture Evidence
Adjacent compare_process_captures failures still lost their reason.
Over MCP, Evidence from another workflow, tampered Evidence and an
invalid capture result all became "Evidence is invalid or has changed".
The CLI said only "Capture evidence is malformed".
Parse each side separately. Evidence from another workflow is an input
issue on left or right that names the operation and predicate it has.
Tampered Evidence and invalid capture results stay integrity
mismatches, and their message now carries the validation detail.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(managed): keep caller-input details in managed workflow errors
The managed workflows wrapped every rejected input as
`new AnalysisInputError(op, { cause })` with no issues. The projected
result said "Correct the listed arguments and retry" and listed nothing:
- `rea import-managed-reconstruction`,
`rea project-managed-application-graph` and
`rea verify-managed-native-boundaries` lost every schema issue for
malformed JSON;
- domain checks such as a reconstruction text hash mismatch, or Evidence
from the wrong operation, lost their message on CLI and MCP;
- `rea compare-managed-members` with a missing file did not say which
path failed.
Add managedInputError, which projects Zod issues with their paths and
keeps a domain TypeError's message, as the JavaScript workflows already
do. Return the BinaryTargetError from an unopenable comparison path
unchanged, with details.path, as every other path-based command does.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(managed): keep nested Evidence schema issues off request paths
A ZodError raised after the request parsed comes from a nested
Evidence result, so its issue paths are relative to that result.
Projecting it against the whole request reported, for example,
artifact.sha256 as a missing request field. Keep request-relative
projection for the raw request's schema failures. Report nested schema
failures as root issues whose message names the nested location.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(managed): type the malformed nested Evidence fixture
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(workflows): keep input details in the remaining workflow services
Inventory projection (project_android_application_graph and
project_apple_application_graph), build_reconstruction_obligation_ledger
and evaluate_reconstruction_coverage had the same issue-less
`AnalysisInputError({ cause })` as the managed services. For example,
evaluating coverage for an unknown boundary returned "Correct the listed
arguments" with no issues, dropping "Unknown reconstruction boundary: X"
on CLI and MCP.
Move the helpers to src/application/workflowInputError.ts and use them
for these services: requestInputError for raw request schema failures,
workflowInputError for nested schema and domain constraint failures.
Classification is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Retain complete authenticated application Evidence when the repeated MCP representation exceeds the default receive frame or Node string limit. Return a typed transport constraint, keep useful followup queries available, and stream full canonical bundles to atomic destinations.
Verify complete Notion and Obsidian workflows on Linux and native Windows, full CLI/Evidence parity, and real large Chrome screenshot recovery. Preserve complete success contracts and caller-selected evidence paths.
Refs #1065, #1053.
A malformed bundle that merely has an evidence_id field was described
as one Evidence record. Use the single-record message only when the
input matches the Evidence record schema; otherwise report the bundle
schema failure.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Reading a nonexistent Evidence bundle or snapshot, or exporting into a
directory that does not exist, failed as reason "io" with "Evidence
file could not be accessed. Check file permissions", and with no path.
In `rea compare` the caller could not tell which of the two files
failed.
Map ENOENT and ENOTDIR to a new "missing" reason with read- and
write-specific advice. Every EvidenceFileError now carries the resolved
caller path in its projected details. Permission and other I/O failures
keep the existing advice.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(process): stop failing captures on tokens macOS cannot expose
On a busy macOS 26 host the residual process sweep fails every capture:
KERN_PROCARGS2 omits the environment of Apple platform binaries for any
caller and fails with EINVAL for processes owned by another account, so an
unrelated process created after the capture baseline escalates to a hard
cleanup failure with cleanup_incomplete.
Report environment withholding distinctly from the Darwin token reader by
reading the platform-binary flag through csops, carry the effective uid in
the process table, and re-read unresolved tokens with a bounded retry plus a
final liveness confirmation so transient exec and exit states settle.
Candidates whose token the OS provably withholds, foreign-uid candidates,
and candidates that exited during validation are now recorded as unverified
processes instead of failing an otherwise clean capture or being signaled.
Related candidates and every other unreadable candidate stay fail-closed.
* test(process): align ownership lanes with the uid-aware process table
The Darwin sweep reads the effective uid through ps, so the offline
verification lanes' syscall allowlists must accept the new argv in
verify-binary-layout and verify-evm-interface. The foreign-uid ownership
test requires POSIX process uids; skip it where the platform has none
(the curated Windows lane).
* fix(verify): recognize UID-bearing ownership process tables
---------
Co-authored-by: N0zoM1z0 <1705004398@qq.com>
evidence-import, evidence-export, compare and the MCP bundle tools gave
every rejected file the same message: "Evidence is invalid or has
changed. Recreate or re-import it". That included a single Evidence
record, which is what every command prints, so `rea compare a.json
b.json` on two command results looked like tampering.
Report a single record as such and point to export_evidence_bundle.
Locate the first schema issue, and keep thrown constraint messages such
as a semantic identifier mismatch. EvidenceIntegrityError now accepts
the userMessage option that other typed errors already use.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A symlink or directory named Assets.car is skipped by the
non-following directory reader. When no regular catalog remains, fail
with the skipped paths instead of reporting an empty inventory. When
catalogs were read, list the skipped entries in the limitations. The
absence limitation now says that symlinks are not followed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
inspect_asset_catalog failed with an "unavailable" artifact error for an
app bundle that has no compiled Assets.car. Its advice was "Check
artifact support and required tools", which does not apply: assetutil
was present, and the absence of a catalog is itself an observation.
decode_interface_builder already returns an empty graph for an app
without NIBs.
Return the empty page with an explicit limitation. Interface Builder
resource keys are still decoded and reported as unmatched.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
inspect_keyed_archive reported caller selection failures as artifact
faults. A missing named root became "Artifact is invalid or has changed.
Get a fresh copy" and lost its reason, and a missing, non-canonical, or
directory archive path became "unsafe or conflicting internal path ...
correct the artifact".
Report these as invalid_request issues on `path` or `root`, listing the
available $top roots or the active plist basename. Decode failures keep
the decoder's reason, and a compiled NIBArchive points to
decode_interface_builder.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(domain): preserve prototype-named JSON members through Evidence boundaries
Zod's z.json() record projection drops own __proto__ members as a
prototype-pollution guard, so provider replies carrying them lost evidence
before hashing/serialization and CLI/MCP output. The shared schema now
validates with z.json() as before, then rebuilds the value keeping
prototype-named own members as ordinary data properties (__proto__ via
Object.defineProperty, so the prototype is never mutated).
parseHarJson no longer rejects __proto__ members; the explicit unsupported
case from #992 is replaced with a preservation regression, and the
inspect_web_network_capture contract text is updated to match.
Fixes#1036
* fix(domain): bound JSON depth and oversized payloads at parsing boundaries
jsonValueSchema re-walked hostile input through z.json()'s recursive core, so a deeply nested payload from an MCP tool argument, inline evidence, provider result, or keyed archive threw RangeError: Maximum call stack size exceeded instead of a validation issue. The schema now runs an iterative depth guard first and keeps z.json() for full JSON semantics behind it, which also turns cyclic values into validation errors. The 1000-edge limit stays far below the range where the recursive core overflows its stack (measured between 2000 and 3000 on Node 24) while leaving room for legitimately deep target-derived trees such as compiled NIB view hierarchies, and the guard narrows with a typed record check instead of any or unchecked casts. The non-recursive JSON Schema advertisement stays intact on both the public and standard projection paths. Two payload sizes get explicit budgets tied to their producer constraints: web text artifacts reject beyond the 16 MiB source budget the web tooling already advertises, and native UI screenshot base64 beyond the 64 MiB helper output budget. Regenerated product catalog, skill catalog digest, and completion-ledger manifest follow the schema-boundary change.
* fix(validation): preserve depth diagnostics across process inputs
---------
Co-authored-by: DeryFerd <deryferdikao125@gmail.com>
Co-authored-by: N0zoM1z0 <1705004398@qq.com>
* fix(domain): bound JSON depth and oversized payloads at parsing boundaries
jsonValueSchema re-walked hostile input through z.json()'s recursive core, so a deeply nested payload from an MCP tool argument, inline evidence, provider result, or keyed archive threw RangeError: Maximum call stack size exceeded instead of a validation issue. The schema now runs an iterative depth guard first and keeps z.json() for full JSON semantics behind it, which also turns cyclic values into validation errors. The 1000-edge limit stays far below the range where the recursive core overflows its stack (measured between 2000 and 3000 on Node 24) while leaving room for legitimately deep target-derived trees such as compiled NIB view hierarchies, and the guard narrows with a typed record check instead of any or unchecked casts. The non-recursive JSON Schema advertisement stays intact on both the public and standard projection paths. Two payload sizes get explicit budgets tied to their producer constraints: web text artifacts reject beyond the 16 MiB source budget the web tooling already advertises, and native UI screenshot base64 beyond the 64 MiB helper output budget. Regenerated product catalog, skill catalog digest, and completion-ledger manifest follow the schema-boundary change.
* fix(validation): preserve depth diagnostics across process inputs
---------
Co-authored-by: N0zoM1z0 <1705004398@qq.com>
* chore(release/6.0.0): release rea-agents 5.1.0
* chore: normalize release product catalog
* chore(release): correct 6.0.0 candidate and migration notes
Keep the recorded application checkpoint unchanged. The upstream history walk
stops at the merged side-branch release commit and omits older unreleased main
commits; restore the missing notes and absolute-path migration instructions.
Set all release metadata to the agreed SemVer 6.0.0.
Assisted-by: Sakuya (gensokyo-skills:sakuya-checkpointed-execution)
---------
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: N0zoM1z0 <1705004398@qq.com>
The MCP SDK reconverts every registered Zod schema on each tools/list.
With 139 tools that is ~45 ms of Zod JSON Schema work per listing, for a
catalog that cannot change after registration. Memoize the advertised
input and output projections per target on the registration clones.
Listings are byte-identical; repeated tools/list drops from ~62 ms to
~18 ms on the stdio server.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* feat(native): inspect recorded crashes through pwntools and pwndbg
* fix(native): bind core note semantics to raw owner bytes
* fix(native): validate recorded core ABI and evidence scalars
* fix(native): require complete recorded note interpretations
* fix(native): bind physical core notes to original headers
main now carries #1037 (skill version 30, reworded native binary row):
- README_ja/ar/ko/zh: take main's tables and count
trace_dylib_resolution in the artifact graph row (5 -> 6).
- Skill: move to version 31.
- Regenerate the product catalog, conformance manifest, skill digest and
package metadata (137 tools).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
main now carries #964 (skill version 29, macOS native row 7 -> 8):
- README_ja/ar/ko/zh: take main's tables and count
trace_dylib_resolution in the artifact graph row (5 -> 6).
- Skill: move to version 30.
- Regenerate the product catalog, conformance manifest, skill digest and
package metadata (136 tools).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Resolve conflicts against main:
- package.json: ship both the LLDB tracer bridge and main's process run
token reader sources.
- README.md: take main's simplified English README, which no longer has
the tool-family table.
- README_ja.md: take main's reworded table and count
observe_native_calls in the macOS native row (7 -> 8).
- Skill: the references gain the observe_native_calls paragraph, so the
version moves to 29 (main already uses 28).
- Regenerate the product catalog, conformance manifest, skill digest and
package metadata (135 tools). The MCP tool catalog is now built rather
than committed (#997), so it is removed here.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Resolve conflicts against main:
- README.md: take main's simplified English README, which no longer has
the tool-family table.
- README_ja.md: take main's reworded table and count
trace_dylib_resolution in the artifact graph row (5 -> 6).
- Skill: the references gain the trace_dylib_resolution paragraph, so
the version moves to 29 (main already uses 28).
- Regenerate the product catalog, conformance manifest, skill digest and
package metadata (135 tools). The MCP tool catalog is now built rather
than committed (#997), so it is removed here.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(mcp): advertise JSON values without recursive schemas
`z.json()` projects a self-referential `$defs/__schema0` entry. Twenty
tool input schemas and 123 output schemas carried it, and clients or
model APIs that reject recursive JSON Schemas refuse the whole tool
list. Every JSON value satisfies `{}`, so the shared JSON value schema
now advertises that. Runtime parsing still admits only JSON values.
Refs #919
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(mcp): keep advertised input schemas within model function-calling limits
Some model providers reject the whole tool list when any input schema
contains a recursive $ref or nests deeper than 10 levels (fixes#919).
Sanitize the advertised wire projection: break cyclic $refs, bound nesting
depth, and drop unreachable $defs. Canonical Zod contracts are untouched.
---------
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: sx4im <sx4im@users.noreply.github.com>
Co-authored-by: N0zoM1z0 <1705004398@qq.com>
`z.json()` projects a self-referential `$defs/__schema0` entry. Twenty
tool input schemas and 123 output schemas carried it, and clients or
model APIs that reject recursive JSON Schemas refuse the whole tool
list. Every JSON value satisfies `{}`, so the shared JSON value schema
now advertises that. Runtime parsing still admits only JSON values.
Refs #919
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: N0zoM1z0 <1705004398@qq.com>
Alan Vo
Tahsin Arafat
Kaoru0822
Rudy Mizrahi Celekli
Ysopwnable
morluto
muhammad-luay
Saim Shafique
Dery Ferdika
ASVLCII
cevieiras
N0zoM1z0
github-actions[bot]
Alex