Validate engine output in the worker and transport output at reply admission. Have the typed function projector consume admitted producer facts in the provider, removing the fused parse/project facade and discarded worker projection.
Preserve duplicate order, missing inferences and raw metadata. Keep malformed producer checks at the schema and cover the provider handoff with source identity, diagnostics and cleanup assertions.
Preserve boundary facts and uncertainty through their consumers. Add focused production-boundary regressions and update affected documentation.
Fixes#6780
The EVM interface failure translator grouped the stable reader's input
limit with path and format refusals, so a carrier over the fixed 4 MiB
input allowance surfaced as a non-retryable invalid_request whose only
record of the allowance was issue text pointing at the input schema.
The same refusal now returns a non-retryable resource_constraint with
resource file-size and reported_limits.maximum_input_bytes, following
the WABT input translator, while path and format reader refusals stay
input errors. The reply-size write failure keeps its own branch.
Fixes#6053
* 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>
Construct integer formats directly while preserving bounds, refinements, metadata and current shared schema owners. Keep Incur-dependent CLI numeric coercion declarations unchanged.
Verify canonical advertisements, 126 numeric boundary cases, 974 actual CLI parser cases, broad source tests and real public-artifact SDK workflows. Record 72 normal tasks and 12 separate owned-inspector heap diagnostics with whole-family resource bounds; distinguish SDK spans and asymmetric audit overhead. Generate numerical report sections from receipts and reject stale tables through documentation conformance checks.
Closes#6236. Current-head CI and Greptile pass; actionable review findings are addressed. Codex review was unavailable due repository usage limits.
cborg's rejectDuplicateMapKeys compares Map keys by identity, and each byte-string key decodes to its own Uint8Array, so a terminal metadata map with the same byte key twice was reported as decoded. Compare the root map's scalar keys by kind and value and report a repeat as not-decoded, like repeated text keys.
The hex carrier trimmed outer ASCII whitespace with a global
start|end-anchored regex. For an invalid carrier with a long interior
whitespace run, the end branch was retried at every position of the run,
so rejection took quadratic time (about 18 s for a 200 KB run) and a few
hundred KB could exhaust the worker deadline instead of returning the
documented invalid-format error. Trim with index scans over the same
ASCII whitespace set.
Co-authored-by: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
* fix(engines): keep format ceilings fixed and recover the EVMole deadline
Promote the owned EVMole worker wall clock to timeout_ms and name it in
recovery. Fixed CPU, byte, and address-space refusals in this theme say
the adjustable deadline does not raise them.
Refs #5910
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(engines): recover the allowance that refused the request
Model capacity recovery follows a structured limit instead of message text,
and the shared header ceiling no longer names Safetensors. Firmware signal
errors keep captured tool output, oversized reports do not suggest an
intermittent rerun, and NativeAOT hydration budget exhaustion explains the
fixed parser allowance on the partial report.
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>
Malformed DOS MZ, raw-binary entry, EVM limit, and Ghidra handshake
cases only checked Result.ok. Name the producer error, or the admitted
language and compiler, so a different failure cannot keep the test green.
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
Share asynchronous executable selection between discovery and execution. Recheck it for default Linux x64 sessions, preserve caller-injected service policy and propagate cancellation through pending probes.
Refs #5086. Focused owner, inventory and MCP controls passed; Linux-specific discovery and execution cases were skipped on Darwin arm64 and remain unverified.
Copy the exact opaque reason already recorded by OwnedCommand rather than discarding it at EVM failure conversion. Keep physical, timeout and incomplete cleanup outcomes distinct.
Validation: direct compiled probe and eight source controls; four reason controls failed before the fix. Twenty-one overlapping service/adapter checks, check:fast and docs:check passed. No real Linux x64 worker cancellation or public delivery exercised. Refs #4686.
Share validated interface Evidence assembly across success, cleanup failure and late cancellation. Preserve exact admission and completion abort reasons.
Validation: 26 service, MCP and host-refusal checks passed; check:fast passed. Thirty Linux x64 provider checks skipped on Darwin arm64, including expanded root-removal controls. Linux real-worker and public SIGINT reproduction remain unverified; no performance measurement. Refs #4683.
Refs #4373, #4455, #4457, #4458.
Use canonical artifact ownership in Go, Rust, SQLite and PE readers and their failure consumers. Admit PE read budgets before allocation and ignore zero-extent sections in interval ownership. Close Rust resources through CLI and MCP lifecycles.
Retain a worker owner transferred through the callback when a failure has no replacement owner. Give captured streams one canonical owner while preserving unrelated diagnostics and partial Evidence. Linux retained-cleanup coverage requires CI.
Preserve parsed worker facts and captured diagnostics when owned worker or snapshot cleanup fails, and retain the owner for retry. Keep primary malformed-reply errors in the cleanup chain.
Validation: build, typecheck and static checks passed. Linux x64 worker regressions added; unavailable on this macOS host.
Keep failed stable-read handles and WABT runtime roots under retryable resource scopes. Preserve cancellation causes and primary errors, close the same service instances bound to MCP tools, and consolidate CLI operation/cleanup projection.
Validation: fast typecheck and lint, cached runtime build, docs checks, SDK schema validation, focused service and real-file cleanup tests, CLI cleanup regressions, and MCP lifecycle/shutdown workflows. Provider process fixtures simulate engine output; real WABT and Linux EVM execution are not established by this pass.
Narrow WASM and EVM execution ports to file subjects and preserve supplied subject types at the shared execution factory. Authenticate format with path and digest before recording or projecting Evidence, using the existing WASM identity owner in both consumers.
Validation: focused WASM/EVM service and public MCP workflows passed, including malformed provider subjects and self-contained retained Evidence with contradictory format. Global typecheck/lint and documentation checks passed. Real WABT/Evmole execution and the full suite were not run.
Retain the original owned-command cleanup failure and one captured-output copy while preserving cleanup resources, truncation and retry semantics.
Verify socket release after cancelled CDP cleanup while retaining the independent detach regression.
Validation: 167 tests across 13 files, all eight local gates, real small Chrome/EVM CLI and pinned-SDK MCP controls, and 20/20 exact-head CI checks.
Host and tool-configuration refusals from the pwntools, EVMole, firmware,
mitmproxy, Wakaru, LLDB and xcrun adapters carried no userMessage, so CLI
and MCP callers saw the generic "This analysis is unavailable for the
current target. Choose another analysis or target." as both message and
remediation. Choosing another target cannot satisfy a Linux-only profile
or an unset REA_*_COMMAND. The actual requirement appeared only in
details.reason.
Pass each refusal's reason as its userMessage, as the adapters' other
configuration refusals (EVMole limits, JADX, pwndbg) already do.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
morluto