fix(cluster): reject truncated command payloads with MALFORMED_COMMAND

[RobinBol]

Summary

Hardens command-argument and attribute-record parsing against truncated wire input. Without these checks, @athombv/data-types silently zero-fills missing bytes, producing a plausible-looking parse result that:

• For IAS Zone zoneStatusChangeNotification and attribute reports of zoneStatus: reads as zoneStatus.alarm1 === false (i.e. "all clear" while truncated frames look indistinguishable from a real "no alarm" report at the consumer layer).
• For any status-bearing response: reads as status === 'SUCCESS' (because the zero value is what every ZCL status enum maps to SUCCESS).

This is a robustness / defense-in-depth change. The framing is correctness-first: safety-critical cluster handlers should never observe args from a payload too short to populate them, and "downstream code cannot tell whether the device reported a valid false value or whether the parser invented it from missing bytes" is a trust-boundary problem worth closing.

Changes

1. Command arg parsing - length precheck + try/catch

Cluster.handleFrame (Cluster.js:780) and BoundCluster.handleFrame (BoundCluster.js:312) now route command-arg parsing through a new helper, lib/util/parseCommandArgs.js:

• Payload shorter than Math.abs(command.args.length) is rejected before the parser runs.
• Payload that throws mid-parse (e.g. an octstr length prefix overruns the remaining buffer) is caught and converted to MALFORMED_COMMAND instead of leaking a RangeError as a generic FAILURE.
• Endpoint.handleFrame already converts thrown ZCLErrors into proper ZCL default-response frames back to the sender, so no surrounding wiring changes are needed.

2. Attribute report (reportAttributes) value-length validation

ZCLAttributeDataRecord.fromBuf in zclFrames.js now validates that enough bytes remain for the value field before calling DataType.fromBuffer. Without this check, a reportAttributes frame carrying a valid attribute header but a truncated value bypassed the command-arg precheck entirely (because reportAttributes declares args as a raw buffer) and re-introduced the same silent zero-fill at the inner record layer. Concretely, a zoneStatus (map16) attribute report missing its value bytes parsed as an all-bits-false bitmap, emitted attr.zoneStatus, and reached IAS-Zone drivers as "all alarms clear".

On insufficient bytes, throws ZCLError('MALFORMED_COMMAND'). Variable-length value fields are not covered by this check (see "Out of scope" below).

3. Tighter encodeMissingFieldsBehavior: 'skip' exemption

The initial precheck fully exempted 'skip' commands (OTA) because they allow trailing fields to be omitted. That was broader than needed: every 'skip' command in the codebase has a mandatory leading discriminator (status / fieldControl / payloadType) whose presence determines which subsequent fields are valid. Without that byte, the parse is meaningless.

The exemption now requires at least the first field's byte width. This preserves the legitimate "omit trailing optional fields" path (e.g. queryNextImageRequest without hardwareVersion) while rejecting empty / sub-leading-byte payloads.

Test plan

All new tests written TDD-style (red first, then implementation).

• Command path - truncated zoneStatusChangeNotification, truncated zoneEnrollResponse:
  • Handlers not invoked.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed equivalents still invoke the handler with the expected args (regression).
• Attribute report path - truncated zoneStatus (map16) value:
  • attr.zoneStatus event does not fire.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed zoneStatus report still fires attr.zoneStatus with alarm1 === true (regression).
• 'skip' exemption - empty queryNextImageRequest:
  • Rejected with MALFORMED_COMMAND.
  • 9-byte queryNextImageRequest without trailing hardwareVersion still parses successfully and reaches the handler with the expected manufacturerCode / imageType (regression - preserves existing OTA tolerance).
• Full suite: 86/86 passing (76 pre-existing + 10 new). No regressions.
• Lint clean.

Out of scope

This PR is consumer-side hardening. The deeper fix sits in @athombv/data-types and includes:

• Fixed-width reads (uintFromBuf, enumFromBuf, dataFromBuf) failing on insufficient bytes instead of returning 0 / a truncated buffer.
• Variable-length reads (bufferFromBuf) failing instead of returning a truncated buffer with isPartial = true (an undocumented flag no consumer checks - the source even contains a // TODO: FIXME next to it).
• Bitmap slices not aliasing the source buffer (TOCTOU hazard if a radio driver reuses receive buffers).
• Struct constructor rejecting prototype-chain keys like constructor as field names.

Two of those (Bitmap defensive copy, Struct prototype-chain guard) are already in flight upstream at athombv/node-data-types#44. The *FromBuf truncation behavior is the breaking change that needs a separate discussion (either a strict opt-in flag or a major-version flip).

As a consequence, this PR does not close:

• Variable-length attribute / command field overrun via bufferFromBuf's isPartial path. The try/catch in parseCommandArgs catches nothing here because nothing throws.
• Attribute reports carrying a variable-length value field (e.g. octstr-typed attributes) where the length prefix overruns the remaining buffer.

These can be closed either upstream (cleanest) or by recursing parsed args downstream to reject any Buffer with isPartial === true (less clean, but unblocks without coordination). Tracked separately.

[Copilot]

Pull request overview

This PR hardens inbound ZCL command argument parsing so truncated/unparseable command payloads are rejected with ZCLError('MALFORMED_COMMAND') instead of being silently parsed (often as all-zero “SUCCESS”/“all-clear” values). This leverages existing Endpoint.handleFrame behavior to automatically emit a proper ZCL Default Response without additional wiring.

Changes:

• Added a shared parseCommandArgs() helper that performs a minimum-length precheck (when applicable) and converts parse-time exceptions into MALFORMED_COMMAND.
• Updated Cluster.handleFrame and BoundCluster.handleFrame to use the new helper instead of calling command.args.fromBuffer(...) directly.
• Added tests to ensure truncated IAS Zone frames don’t invoke handlers and do produce MALFORMED_COMMAND Default Responses (with regressions for well-formed frames).

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
test/testMalformedFrames.js	Adds regression + hardening tests verifying truncated payloads yield MALFORMED_COMMAND and do not call handlers.
lib/util/parseCommandArgs.js	Introduces defensive argument parsing helper with length precheck + parse error normalization.
lib/Cluster.js	Routes inbound cluster command arg parsing through parseCommandArgs().
lib/BoundCluster.js	Routes inbound bound-cluster command arg parsing through parseCommandArgs().

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[RobinBol]

Summary

Hardens command-argument and attribute-record parsing against truncated wire input. Without these checks, @athombv/data-types silently zero-fills missing bytes, producing a plausible-looking parse result that:

• For IAS Zone zoneStatusChangeNotification and attribute reports of zoneStatus: reads as zoneStatus.alarm1 === false (i.e. "all clear" while truncated frames look indistinguishable from a real "no alarm" report at the consumer layer).
• For any status-bearing response: reads as status === 'SUCCESS' (because the zero value is what every ZCL status enum maps to SUCCESS).

This is a robustness / defense-in-depth change. The framing is correctness-first: safety-critical cluster handlers should never observe args from a payload too short to populate them, and "downstream code cannot tell whether the device reported a valid false value or whether the parser invented it from missing bytes" is a trust-boundary problem worth closing.

Changes

1. Command arg parsing - length precheck + try/catch

Cluster.handleFrame (Cluster.js:780) and BoundCluster.handleFrame (BoundCluster.js:312) now route command-arg parsing through a new helper, lib/util/parseCommandArgs.js:

• Payload shorter than Math.abs(command.args.length) is rejected before the parser runs.
• Payload that throws mid-parse (e.g. an octstr length prefix overruns the remaining buffer) is caught and converted to MALFORMED_COMMAND instead of leaking a RangeError as a generic FAILURE.
• Endpoint.handleFrame already converts thrown ZCLErrors into proper ZCL default-response frames back to the sender, so no surrounding wiring changes are needed.

2. Attribute report (reportAttributes) value-length validation

ZCLAttributeDataRecord.fromBuf in zclFrames.js now validates that enough bytes remain for the value field before calling DataType.fromBuffer. Without this check, a reportAttributes frame carrying a valid attribute header but a truncated value bypassed the command-arg precheck entirely (because reportAttributes declares args as a raw buffer) and re-introduced the same silent zero-fill at the inner record layer. Concretely, a zoneStatus (map16) attribute report missing its value bytes parsed as an all-bits-false bitmap, emitted attr.zoneStatus, and reached IAS-Zone drivers as "all alarms clear".

On insufficient bytes, throws ZCLError('MALFORMED_COMMAND'). Variable-length value fields are not covered by this check (see "Out of scope" below).

3. Tighter encodeMissingFieldsBehavior: 'skip' exemption

The initial precheck fully exempted 'skip' commands (OTA) because they allow trailing fields to be omitted. That was broader than needed: every 'skip' command in the codebase has a mandatory leading discriminator (status / fieldControl / payloadType) whose presence determines which subsequent fields are valid. Without that byte, the parse is meaningless.

The exemption now requires at least the first field's byte width. This preserves the legitimate "omit trailing optional fields" path (e.g. queryNextImageRequest without hardwareVersion) while rejecting empty / sub-leading-byte payloads.

Test plan

All new tests written TDD-style (red first, then implementation).

• Command path - truncated zoneStatusChangeNotification, truncated zoneEnrollResponse:
  • Handlers not invoked.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed equivalents still invoke the handler with the expected args (regression).
• Attribute report path - truncated zoneStatus (map16) value:
  • attr.zoneStatus event does not fire.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed zoneStatus report still fires attr.zoneStatus with alarm1 === true (regression).
• 'skip' exemption - empty queryNextImageRequest:
  • Rejected with MALFORMED_COMMAND.
  • 9-byte queryNextImageRequest without trailing hardwareVersion still parses successfully and reaches the handler with the expected manufacturerCode / imageType (regression - preserves existing OTA tolerance).
• Full suite: 86/86 passing (76 pre-existing + 10 new). No regressions.
• Lint clean.

Out of scope

This PR is consumer-side hardening. The deeper fix sits in @athombv/data-types and includes:

• Fixed-width reads (uintFromBuf, enumFromBuf, dataFromBuf) failing on insufficient bytes instead of returning 0 / a truncated buffer.
• Variable-length reads (bufferFromBuf) failing instead of returning a truncated buffer with isPartial = true (an undocumented flag no consumer checks - the source even contains a // TODO: FIXME next to it).
• Bitmap slices not aliasing the source buffer (TOCTOU hazard if a radio driver reuses receive buffers).
• Struct constructor rejecting prototype-chain keys like constructor as field names.

Two of those (Bitmap defensive copy, Struct prototype-chain guard) are already in flight upstream at athombv/node-data-types#44. The *FromBuf truncation behavior is the breaking change that needs a separate discussion (either a strict opt-in flag or a major-version flip).

As a consequence, this PR does not close:

• Variable-length attribute / command field overrun via bufferFromBuf's isPartial path. The try/catch in parseCommandArgs catches nothing here because nothing throws.
• Attribute reports carrying a variable-length value field (e.g. octstr-typed attributes) where the length prefix overruns the remaining buffer.

These can be closed either upstream (cleanest) or by recursing parsed args downstream to reject any Buffer with isPartial === true (less clean, but unblocks without coordination). Tracked separately.

[Copilot]

Pull request overview

This PR hardens inbound ZCL command argument parsing so truncated/unparseable command payloads are rejected with ZCLError('MALFORMED_COMMAND') instead of being silently parsed (often as all-zero “SUCCESS”/“all-clear” values). This leverages existing Endpoint.handleFrame behavior to automatically emit a proper ZCL Default Response without additional wiring.

Changes:

• Added a shared parseCommandArgs() helper that performs a minimum-length precheck (when applicable) and converts parse-time exceptions into MALFORMED_COMMAND.
• Updated Cluster.handleFrame and BoundCluster.handleFrame to use the new helper instead of calling command.args.fromBuffer(...) directly.
• Added tests to ensure truncated IAS Zone frames don’t invoke handlers and do produce MALFORMED_COMMAND Default Responses (with regressions for well-formed frames).

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
test/testMalformedFrames.js	Adds regression + hardening tests verifying truncated payloads yield MALFORMED_COMMAND and do not call handlers.
lib/util/parseCommandArgs.js	Introduces defensive argument parsing helper with length precheck + parse error normalization.
lib/Cluster.js	Routes inbound cluster command arg parsing through parseCommandArgs().
lib/BoundCluster.js	Routes inbound bound-cluster command arg parsing through parseCommandArgs().

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[RobinBol]

Summary

Hardens command-argument and attribute-record parsing against truncated wire input. Without these checks, @athombv/data-types silently zero-fills missing bytes, producing a plausible-looking parse result that:

• For IAS Zone zoneStatusChangeNotification and attribute reports of zoneStatus: reads as zoneStatus.alarm1 === false (i.e. "all clear" while truncated frames look indistinguishable from a real "no alarm" report at the consumer layer).
• For any status-bearing response: reads as status === 'SUCCESS' (because the zero value is what every ZCL status enum maps to SUCCESS).

This is a robustness / defense-in-depth change. The framing is correctness-first: safety-critical cluster handlers should never observe args from a payload too short to populate them, and "downstream code cannot tell whether the device reported a valid false value or whether the parser invented it from missing bytes" is a trust-boundary problem worth closing.

Changes

1. Command arg parsing - length precheck + try/catch

Cluster.handleFrame (Cluster.js:780) and BoundCluster.handleFrame (BoundCluster.js:312) now route command-arg parsing through a new helper, lib/util/parseCommandArgs.js:

• Payload shorter than Math.abs(command.args.length) is rejected before the parser runs.
• Payload that throws mid-parse (e.g. an octstr length prefix overruns the remaining buffer) is caught and converted to MALFORMED_COMMAND instead of leaking a RangeError as a generic FAILURE.
• Endpoint.handleFrame already converts thrown ZCLErrors into proper ZCL default-response frames back to the sender, so no surrounding wiring changes are needed.

2. Attribute report (reportAttributes) value-length validation

ZCLAttributeDataRecord.fromBuf in zclFrames.js now validates that enough bytes remain for the value field before calling DataType.fromBuffer. Without this check, a reportAttributes frame carrying a valid attribute header but a truncated value bypassed the command-arg precheck entirely (because reportAttributes declares args as a raw buffer) and re-introduced the same silent zero-fill at the inner record layer. Concretely, a zoneStatus (map16) attribute report missing its value bytes parsed as an all-bits-false bitmap, emitted attr.zoneStatus, and reached IAS-Zone drivers as "all alarms clear".

On insufficient bytes, throws ZCLError('MALFORMED_COMMAND'). Variable-length value fields are not covered by this check (see "Out of scope" below).

3. Tighter encodeMissingFieldsBehavior: 'skip' exemption

The initial precheck fully exempted 'skip' commands (OTA) because they allow trailing fields to be omitted. That was broader than needed: every 'skip' command in the codebase has a mandatory leading discriminator (status / fieldControl / payloadType) whose presence determines which subsequent fields are valid. Without that byte, the parse is meaningless.

The exemption now requires at least the first field's byte width. This preserves the legitimate "omit trailing optional fields" path (e.g. queryNextImageRequest without hardwareVersion) while rejecting empty / sub-leading-byte payloads.

Test plan

All new tests written TDD-style (red first, then implementation).

• Command path - truncated zoneStatusChangeNotification, truncated zoneEnrollResponse:
  • Handlers not invoked.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed equivalents still invoke the handler with the expected args (regression).
• Attribute report path - truncated zoneStatus (map16) value:
  • attr.zoneStatus event does not fire.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed zoneStatus report still fires attr.zoneStatus with alarm1 === true (regression).
• 'skip' exemption - empty queryNextImageRequest:
  • Rejected with MALFORMED_COMMAND.
  • 9-byte queryNextImageRequest without trailing hardwareVersion still parses successfully and reaches the handler with the expected manufacturerCode / imageType (regression - preserves existing OTA tolerance).
• Full suite: 86/86 passing (76 pre-existing + 10 new). No regressions.
• Lint clean.

Out of scope

This PR is consumer-side hardening. The deeper fix sits in @athombv/data-types and includes:

• Fixed-width reads (uintFromBuf, enumFromBuf, dataFromBuf) failing on insufficient bytes instead of returning 0 / a truncated buffer.
• Variable-length reads (bufferFromBuf) failing instead of returning a truncated buffer with isPartial = true (an undocumented flag no consumer checks - the source even contains a // TODO: FIXME next to it).
• Bitmap slices not aliasing the source buffer (TOCTOU hazard if a radio driver reuses receive buffers).
• Struct constructor rejecting prototype-chain keys like constructor as field names.

Two of those (Bitmap defensive copy, Struct prototype-chain guard) are already in flight upstream at athombv/node-data-types#44. The *FromBuf truncation behavior is the breaking change that needs a separate discussion (either a strict opt-in flag or a major-version flip).

As a consequence, this PR does not close:

• Variable-length attribute / command field overrun via bufferFromBuf's isPartial path. The try/catch in parseCommandArgs catches nothing here because nothing throws.
• Attribute reports carrying a variable-length value field (e.g. octstr-typed attributes) where the length prefix overruns the remaining buffer.

These can be closed either upstream (cleanest) or by recursing parsed args downstream to reject any Buffer with isPartial === true (less clean, but unblocks without coordination). Tracked separately.

[Copilot]

Pull request overview

This PR hardens inbound ZCL command argument parsing so truncated/unparseable command payloads are rejected with ZCLError('MALFORMED_COMMAND') instead of being silently parsed (often as all-zero “SUCCESS”/“all-clear” values). This leverages existing Endpoint.handleFrame behavior to automatically emit a proper ZCL Default Response without additional wiring.

Changes:

• Added a shared parseCommandArgs() helper that performs a minimum-length precheck (when applicable) and converts parse-time exceptions into MALFORMED_COMMAND.
• Updated Cluster.handleFrame and BoundCluster.handleFrame to use the new helper instead of calling command.args.fromBuffer(...) directly.
• Added tests to ensure truncated IAS Zone frames don’t invoke handlers and do produce MALFORMED_COMMAND Default Responses (with regressions for well-formed frames).

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
test/testMalformedFrames.js	Adds regression + hardening tests verifying truncated payloads yield MALFORMED_COMMAND and do not call handlers.
lib/util/parseCommandArgs.js	Introduces defensive argument parsing helper with length precheck + parse error normalization.
lib/Cluster.js	Routes inbound cluster command arg parsing through parseCommandArgs().
lib/BoundCluster.js	Routes inbound bound-cluster command arg parsing through parseCommandArgs().

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[RobinBol]

Summary

Hardens command-argument and attribute-record parsing against truncated wire input. Without these checks, @athombv/data-types silently zero-fills missing bytes, producing a plausible-looking parse result that:

• For IAS Zone zoneStatusChangeNotification and attribute reports of zoneStatus: reads as zoneStatus.alarm1 === false (i.e. "all clear" while truncated frames look indistinguishable from a real "no alarm" report at the consumer layer).
• For any status-bearing response: reads as status === 'SUCCESS' (because the zero value is what every ZCL status enum maps to SUCCESS).

This is a robustness / defense-in-depth change. The framing is correctness-first: safety-critical cluster handlers should never observe args from a payload too short to populate them, and "downstream code cannot tell whether the device reported a valid false value or whether the parser invented it from missing bytes" is a trust-boundary problem worth closing.

Changes

1. Command arg parsing - length precheck + try/catch

Cluster.handleFrame (Cluster.js:780) and BoundCluster.handleFrame (BoundCluster.js:312) now route command-arg parsing through a new helper, lib/util/parseCommandArgs.js:

• Payload shorter than Math.abs(command.args.length) is rejected before the parser runs.
• Payload that throws mid-parse (e.g. an octstr length prefix overruns the remaining buffer) is caught and converted to MALFORMED_COMMAND instead of leaking a RangeError as a generic FAILURE.
• Endpoint.handleFrame already converts thrown ZCLErrors into proper ZCL default-response frames back to the sender, so no surrounding wiring changes are needed.

2. Attribute report (reportAttributes) value-length validation

ZCLAttributeDataRecord.fromBuf in zclFrames.js now validates that enough bytes remain for the value field before calling DataType.fromBuffer. Without this check, a reportAttributes frame carrying a valid attribute header but a truncated value bypassed the command-arg precheck entirely (because reportAttributes declares args as a raw buffer) and re-introduced the same silent zero-fill at the inner record layer. Concretely, a zoneStatus (map16) attribute report missing its value bytes parsed as an all-bits-false bitmap, emitted attr.zoneStatus, and reached IAS-Zone drivers as "all alarms clear".

On insufficient bytes, throws ZCLError('MALFORMED_COMMAND'). Variable-length value fields are not covered by this check (see "Out of scope" below).

3. Tighter encodeMissingFieldsBehavior: 'skip' exemption

The initial precheck fully exempted 'skip' commands (OTA) because they allow trailing fields to be omitted. That was broader than needed: every 'skip' command in the codebase has a mandatory leading discriminator (status / fieldControl / payloadType) whose presence determines which subsequent fields are valid. Without that byte, the parse is meaningless.

The exemption now requires at least the first field's byte width. This preserves the legitimate "omit trailing optional fields" path (e.g. queryNextImageRequest without hardwareVersion) while rejecting empty / sub-leading-byte payloads.

Test plan

All new tests written TDD-style (red first, then implementation).

• Command path - truncated zoneStatusChangeNotification, truncated zoneEnrollResponse:
  • Handlers not invoked.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed equivalents still invoke the handler with the expected args (regression).
• Attribute report path - truncated zoneStatus (map16) value:
  • attr.zoneStatus event does not fire.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed zoneStatus report still fires attr.zoneStatus with alarm1 === true (regression).
• 'skip' exemption - empty queryNextImageRequest:
  • Rejected with MALFORMED_COMMAND.
  • 9-byte queryNextImageRequest without trailing hardwareVersion still parses successfully and reaches the handler with the expected manufacturerCode / imageType (regression - preserves existing OTA tolerance).
• Full suite: 86/86 passing (76 pre-existing + 10 new). No regressions.
• Lint clean.

Out of scope

This PR is consumer-side hardening. The deeper fix sits in @athombv/data-types and includes:

• Fixed-width reads (uintFromBuf, enumFromBuf, dataFromBuf) failing on insufficient bytes instead of returning 0 / a truncated buffer.
• Variable-length reads (bufferFromBuf) failing instead of returning a truncated buffer with isPartial = true (an undocumented flag no consumer checks - the source even contains a // TODO: FIXME next to it).
• Bitmap slices not aliasing the source buffer (TOCTOU hazard if a radio driver reuses receive buffers).
• Struct constructor rejecting prototype-chain keys like constructor as field names.

Two of those (Bitmap defensive copy, Struct prototype-chain guard) are already in flight upstream at athombv/node-data-types#44. The *FromBuf truncation behavior is the breaking change that needs a separate discussion (either a strict opt-in flag or a major-version flip).

As a consequence, this PR does not close:

• Variable-length attribute / command field overrun via bufferFromBuf's isPartial path. The try/catch in parseCommandArgs catches nothing here because nothing throws.
• Attribute reports carrying a variable-length value field (e.g. octstr-typed attributes) where the length prefix overruns the remaining buffer.

These can be closed either upstream (cleanest) or by recursing parsed args downstream to reject any Buffer with isPartial === true (less clean, but unblocks without coordination). Tracked separately.

[Copilot]

Pull request overview

This PR hardens inbound ZCL command argument parsing so truncated/unparseable command payloads are rejected with ZCLError('MALFORMED_COMMAND') instead of being silently parsed (often as all-zero “SUCCESS”/“all-clear” values). This leverages existing Endpoint.handleFrame behavior to automatically emit a proper ZCL Default Response without additional wiring.

Changes:

• Added a shared parseCommandArgs() helper that performs a minimum-length precheck (when applicable) and converts parse-time exceptions into MALFORMED_COMMAND.
• Updated Cluster.handleFrame and BoundCluster.handleFrame to use the new helper instead of calling command.args.fromBuffer(...) directly.
• Added tests to ensure truncated IAS Zone frames don’t invoke handlers and do produce MALFORMED_COMMAND Default Responses (with regressions for well-formed frames).

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
test/testMalformedFrames.js	Adds regression + hardening tests verifying truncated payloads yield MALFORMED_COMMAND and do not call handlers.
lib/util/parseCommandArgs.js	Introduces defensive argument parsing helper with length precheck + parse error normalization.
lib/Cluster.js	Routes inbound cluster command arg parsing through parseCommandArgs().
lib/BoundCluster.js	Routes inbound bound-cluster command arg parsing through parseCommandArgs().

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[RobinBol]

Summary

Hardens command-argument and attribute-record parsing against truncated wire input. Without these checks, @athombv/data-types silently zero-fills missing bytes, producing a plausible-looking parse result that:

• For IAS Zone zoneStatusChangeNotification and attribute reports of zoneStatus: reads as zoneStatus.alarm1 === false (i.e. "all clear" while truncated frames look indistinguishable from a real "no alarm" report at the consumer layer).
• For any status-bearing response: reads as status === 'SUCCESS' (because the zero value is what every ZCL status enum maps to SUCCESS).

This is a robustness / defense-in-depth change. The framing is correctness-first: safety-critical cluster handlers should never observe args from a payload too short to populate them, and "downstream code cannot tell whether the device reported a valid false value or whether the parser invented it from missing bytes" is a trust-boundary problem worth closing.

Changes

1. Command arg parsing - length precheck + try/catch

Cluster.handleFrame (Cluster.js:780) and BoundCluster.handleFrame (BoundCluster.js:312) now route command-arg parsing through a new helper, lib/util/parseCommandArgs.js:

• Payload shorter than Math.abs(command.args.length) is rejected before the parser runs.
• Payload that throws mid-parse (e.g. an octstr length prefix overruns the remaining buffer) is caught and converted to MALFORMED_COMMAND instead of leaking a RangeError as a generic FAILURE.
• Endpoint.handleFrame already converts thrown ZCLErrors into proper ZCL default-response frames back to the sender, so no surrounding wiring changes are needed.

2. Attribute report (reportAttributes) value-length validation

ZCLAttributeDataRecord.fromBuf in zclFrames.js now validates that enough bytes remain for the value field before calling DataType.fromBuffer. Without this check, a reportAttributes frame carrying a valid attribute header but a truncated value bypassed the command-arg precheck entirely (because reportAttributes declares args as a raw buffer) and re-introduced the same silent zero-fill at the inner record layer. Concretely, a zoneStatus (map16) attribute report missing its value bytes parsed as an all-bits-false bitmap, emitted attr.zoneStatus, and reached IAS-Zone drivers as "all alarms clear".

On insufficient bytes, throws ZCLError('MALFORMED_COMMAND'). Variable-length value fields are not covered by this check (see "Out of scope" below).

3. Tighter encodeMissingFieldsBehavior: 'skip' exemption

The initial precheck fully exempted 'skip' commands (OTA) because they allow trailing fields to be omitted. That was broader than needed: every 'skip' command in the codebase has a mandatory leading discriminator (status / fieldControl / payloadType) whose presence determines which subsequent fields are valid. Without that byte, the parse is meaningless.

The exemption now requires at least the first field's byte width. This preserves the legitimate "omit trailing optional fields" path (e.g. queryNextImageRequest without hardwareVersion) while rejecting empty / sub-leading-byte payloads.

Test plan

All new tests written TDD-style (red first, then implementation).

• Command path - truncated zoneStatusChangeNotification, truncated zoneEnrollResponse:
  • Handlers not invoked.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed equivalents still invoke the handler with the expected args (regression).
• Attribute report path - truncated zoneStatus (map16) value:
  • attr.zoneStatus event does not fire.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed zoneStatus report still fires attr.zoneStatus with alarm1 === true (regression).
• 'skip' exemption - empty queryNextImageRequest:
  • Rejected with MALFORMED_COMMAND.
  • 9-byte queryNextImageRequest without trailing hardwareVersion still parses successfully and reaches the handler with the expected manufacturerCode / imageType (regression - preserves existing OTA tolerance).
• Full suite: 86/86 passing (76 pre-existing + 10 new). No regressions.
• Lint clean.

Out of scope

This PR is consumer-side hardening. The deeper fix sits in @athombv/data-types and includes:

• Fixed-width reads (uintFromBuf, enumFromBuf, dataFromBuf) failing on insufficient bytes instead of returning 0 / a truncated buffer.
• Variable-length reads (bufferFromBuf) failing instead of returning a truncated buffer with isPartial = true (an undocumented flag no consumer checks - the source even contains a // TODO: FIXME next to it).
• Bitmap slices not aliasing the source buffer (TOCTOU hazard if a radio driver reuses receive buffers).
• Struct constructor rejecting prototype-chain keys like constructor as field names.

Two of those (Bitmap defensive copy, Struct prototype-chain guard) are already in flight upstream at athombv/node-data-types#44. The *FromBuf truncation behavior is the breaking change that needs a separate discussion (either a strict opt-in flag or a major-version flip).

As a consequence, this PR does not close:

• Variable-length attribute / command field overrun via bufferFromBuf's isPartial path. The try/catch in parseCommandArgs catches nothing here because nothing throws.
• Attribute reports carrying a variable-length value field (e.g. octstr-typed attributes) where the length prefix overruns the remaining buffer.

These can be closed either upstream (cleanest) or by recursing parsed args downstream to reject any Buffer with isPartial === true (less clean, but unblocks without coordination). Tracked separately.

[Copilot]

Pull request overview

This PR hardens inbound ZCL command argument parsing so truncated/unparseable command payloads are rejected with ZCLError('MALFORMED_COMMAND') instead of being silently parsed (often as all-zero “SUCCESS”/“all-clear” values). This leverages existing Endpoint.handleFrame behavior to automatically emit a proper ZCL Default Response without additional wiring.

Changes:

• Added a shared parseCommandArgs() helper that performs a minimum-length precheck (when applicable) and converts parse-time exceptions into MALFORMED_COMMAND.
• Updated Cluster.handleFrame and BoundCluster.handleFrame to use the new helper instead of calling command.args.fromBuffer(...) directly.
• Added tests to ensure truncated IAS Zone frames don’t invoke handlers and do produce MALFORMED_COMMAND Default Responses (with regressions for well-formed frames).

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
test/testMalformedFrames.js	Adds regression + hardening tests verifying truncated payloads yield MALFORMED_COMMAND and do not call handlers.
lib/util/parseCommandArgs.js	Introduces defensive argument parsing helper with length precheck + parse error normalization.
lib/Cluster.js	Routes inbound cluster command arg parsing through parseCommandArgs().
lib/BoundCluster.js	Routes inbound bound-cluster command arg parsing through parseCommandArgs().

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[RobinBol]

Summary

Hardens command-argument and attribute-record parsing against truncated wire input. Without these checks, @athombv/data-types silently zero-fills missing bytes, producing a plausible-looking parse result that:

• For IAS Zone zoneStatusChangeNotification and attribute reports of zoneStatus: reads as zoneStatus.alarm1 === false (i.e. "all clear" while truncated frames look indistinguishable from a real "no alarm" report at the consumer layer).
• For any status-bearing response: reads as status === 'SUCCESS' (because the zero value is what every ZCL status enum maps to SUCCESS).

This is a robustness / defense-in-depth change. The framing is correctness-first: safety-critical cluster handlers should never observe args from a payload too short to populate them, and "downstream code cannot tell whether the device reported a valid false value or whether the parser invented it from missing bytes" is a trust-boundary problem worth closing.

Changes

1. Command arg parsing - length precheck + try/catch

Cluster.handleFrame (Cluster.js:780) and BoundCluster.handleFrame (BoundCluster.js:312) now route command-arg parsing through a new helper, lib/util/parseCommandArgs.js:

• Payload shorter than Math.abs(command.args.length) is rejected before the parser runs.
• Payload that throws mid-parse (e.g. an octstr length prefix overruns the remaining buffer) is caught and converted to MALFORMED_COMMAND instead of leaking a RangeError as a generic FAILURE.
• Endpoint.handleFrame already converts thrown ZCLErrors into proper ZCL default-response frames back to the sender, so no surrounding wiring changes are needed.

2. Attribute report (reportAttributes) value-length validation

ZCLAttributeDataRecord.fromBuf in zclFrames.js now validates that enough bytes remain for the value field before calling DataType.fromBuffer. Without this check, a reportAttributes frame carrying a valid attribute header but a truncated value bypassed the command-arg precheck entirely (because reportAttributes declares args as a raw buffer) and re-introduced the same silent zero-fill at the inner record layer. Concretely, a zoneStatus (map16) attribute report missing its value bytes parsed as an all-bits-false bitmap, emitted attr.zoneStatus, and reached IAS-Zone drivers as "all alarms clear".

On insufficient bytes, throws ZCLError('MALFORMED_COMMAND'). Variable-length value fields are not covered by this check (see "Out of scope" below).

3. Tighter encodeMissingFieldsBehavior: 'skip' exemption

The initial precheck fully exempted 'skip' commands (OTA) because they allow trailing fields to be omitted. That was broader than needed: every 'skip' command in the codebase has a mandatory leading discriminator (status / fieldControl / payloadType) whose presence determines which subsequent fields are valid. Without that byte, the parse is meaningless.

The exemption now requires at least the first field's byte width. This preserves the legitimate "omit trailing optional fields" path (e.g. queryNextImageRequest without hardwareVersion) while rejecting empty / sub-leading-byte payloads.

Test plan

All new tests written TDD-style (red first, then implementation).

• Command path - truncated zoneStatusChangeNotification, truncated zoneEnrollResponse:
  • Handlers not invoked.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed equivalents still invoke the handler with the expected args (regression).
• Attribute report path - truncated zoneStatus (map16) value:
  • attr.zoneStatus event does not fire.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed zoneStatus report still fires attr.zoneStatus with alarm1 === true (regression).
• 'skip' exemption - empty queryNextImageRequest:
  • Rejected with MALFORMED_COMMAND.
  • 9-byte queryNextImageRequest without trailing hardwareVersion still parses successfully and reaches the handler with the expected manufacturerCode / imageType (regression - preserves existing OTA tolerance).
• Full suite: 86/86 passing (76 pre-existing + 10 new). No regressions.
• Lint clean.

Out of scope

This PR is consumer-side hardening. The deeper fix sits in @athombv/data-types and includes:

• Fixed-width reads (uintFromBuf, enumFromBuf, dataFromBuf) failing on insufficient bytes instead of returning 0 / a truncated buffer.
• Variable-length reads (bufferFromBuf) failing instead of returning a truncated buffer with isPartial = true (an undocumented flag no consumer checks - the source even contains a // TODO: FIXME next to it).
• Bitmap slices not aliasing the source buffer (TOCTOU hazard if a radio driver reuses receive buffers).
• Struct constructor rejecting prototype-chain keys like constructor as field names.

Two of those (Bitmap defensive copy, Struct prototype-chain guard) are already in flight upstream at athombv/node-data-types#44. The *FromBuf truncation behavior is the breaking change that needs a separate discussion (either a strict opt-in flag or a major-version flip).

As a consequence, this PR does not close:

• Variable-length attribute / command field overrun via bufferFromBuf's isPartial path. The try/catch in parseCommandArgs catches nothing here because nothing throws.
• Attribute reports carrying a variable-length value field (e.g. octstr-typed attributes) where the length prefix overruns the remaining buffer.

These can be closed either upstream (cleanest) or by recursing parsed args downstream to reject any Buffer with isPartial === true (less clean, but unblocks without coordination). Tracked separately.

[Copilot]

Pull request overview

This PR hardens inbound ZCL command argument parsing so truncated/unparseable command payloads are rejected with ZCLError('MALFORMED_COMMAND') instead of being silently parsed (often as all-zero “SUCCESS”/“all-clear” values). This leverages existing Endpoint.handleFrame behavior to automatically emit a proper ZCL Default Response without additional wiring.

Changes:

• Added a shared parseCommandArgs() helper that performs a minimum-length precheck (when applicable) and converts parse-time exceptions into MALFORMED_COMMAND.
• Updated Cluster.handleFrame and BoundCluster.handleFrame to use the new helper instead of calling command.args.fromBuffer(...) directly.
• Added tests to ensure truncated IAS Zone frames don’t invoke handlers and do produce MALFORMED_COMMAND Default Responses (with regressions for well-formed frames).

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
test/testMalformedFrames.js	Adds regression + hardening tests verifying truncated payloads yield MALFORMED_COMMAND and do not call handlers.
lib/util/parseCommandArgs.js	Introduces defensive argument parsing helper with length precheck + parse error normalization.
lib/Cluster.js	Routes inbound cluster command arg parsing through parseCommandArgs().
lib/BoundCluster.js	Routes inbound bound-cluster command arg parsing through parseCommandArgs().

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[RobinBol]

Summary

Hardens command-argument and attribute-record parsing against truncated wire input. Without these checks, @athombv/data-types silently zero-fills missing bytes, producing a plausible-looking parse result that:

• For IAS Zone zoneStatusChangeNotification and attribute reports of zoneStatus: reads as zoneStatus.alarm1 === false (i.e. "all clear" while truncated frames look indistinguishable from a real "no alarm" report at the consumer layer).
• For any status-bearing response: reads as status === 'SUCCESS' (because the zero value is what every ZCL status enum maps to SUCCESS).

This is a robustness / defense-in-depth change. The framing is correctness-first: safety-critical cluster handlers should never observe args from a payload too short to populate them, and "downstream code cannot tell whether the device reported a valid false value or whether the parser invented it from missing bytes" is a trust-boundary problem worth closing.

Changes

1. Command arg parsing - length precheck + try/catch

Cluster.handleFrame (Cluster.js:780) and BoundCluster.handleFrame (BoundCluster.js:312) now route command-arg parsing through a new helper, lib/util/parseCommandArgs.js:

• Payload shorter than Math.abs(command.args.length) is rejected before the parser runs.
• Payload that throws mid-parse (e.g. an octstr length prefix overruns the remaining buffer) is caught and converted to MALFORMED_COMMAND instead of leaking a RangeError as a generic FAILURE.
• Endpoint.handleFrame already converts thrown ZCLErrors into proper ZCL default-response frames back to the sender, so no surrounding wiring changes are needed.

2. Attribute report (reportAttributes) value-length validation

ZCLAttributeDataRecord.fromBuf in zclFrames.js now validates that enough bytes remain for the value field before calling DataType.fromBuffer. Without this check, a reportAttributes frame carrying a valid attribute header but a truncated value bypassed the command-arg precheck entirely (because reportAttributes declares args as a raw buffer) and re-introduced the same silent zero-fill at the inner record layer. Concretely, a zoneStatus (map16) attribute report missing its value bytes parsed as an all-bits-false bitmap, emitted attr.zoneStatus, and reached IAS-Zone drivers as "all alarms clear".

On insufficient bytes, throws ZCLError('MALFORMED_COMMAND'). Variable-length value fields are not covered by this check (see "Out of scope" below).

3. Tighter encodeMissingFieldsBehavior: 'skip' exemption

The initial precheck fully exempted 'skip' commands (OTA) because they allow trailing fields to be omitted. That was broader than needed: every 'skip' command in the codebase has a mandatory leading discriminator (status / fieldControl / payloadType) whose presence determines which subsequent fields are valid. Without that byte, the parse is meaningless.

The exemption now requires at least the first field's byte width. This preserves the legitimate "omit trailing optional fields" path (e.g. queryNextImageRequest without hardwareVersion) while rejecting empty / sub-leading-byte payloads.

Test plan

All new tests written TDD-style (red first, then implementation).

• Command path - truncated zoneStatusChangeNotification, truncated zoneEnrollResponse:
  • Handlers not invoked.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed equivalents still invoke the handler with the expected args (regression).
• Attribute report path - truncated zoneStatus (map16) value:
  • attr.zoneStatus event does not fire.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed zoneStatus report still fires attr.zoneStatus with alarm1 === true (regression).
• 'skip' exemption - empty queryNextImageRequest:
  • Rejected with MALFORMED_COMMAND.
  • 9-byte queryNextImageRequest without trailing hardwareVersion still parses successfully and reaches the handler with the expected manufacturerCode / imageType (regression - preserves existing OTA tolerance).
• Full suite: 86/86 passing (76 pre-existing + 10 new). No regressions.
• Lint clean.

Out of scope

This PR is consumer-side hardening. The deeper fix sits in @athombv/data-types and includes:

• Fixed-width reads (uintFromBuf, enumFromBuf, dataFromBuf) failing on insufficient bytes instead of returning 0 / a truncated buffer.
• Variable-length reads (bufferFromBuf) failing instead of returning a truncated buffer with isPartial = true (an undocumented flag no consumer checks - the source even contains a // TODO: FIXME next to it).
• Bitmap slices not aliasing the source buffer (TOCTOU hazard if a radio driver reuses receive buffers).
• Struct constructor rejecting prototype-chain keys like constructor as field names.

Two of those (Bitmap defensive copy, Struct prototype-chain guard) are already in flight upstream at athombv/node-data-types#44. The *FromBuf truncation behavior is the breaking change that needs a separate discussion (either a strict opt-in flag or a major-version flip).

As a consequence, this PR does not close:

• Variable-length attribute / command field overrun via bufferFromBuf's isPartial path. The try/catch in parseCommandArgs catches nothing here because nothing throws.
• Attribute reports carrying a variable-length value field (e.g. octstr-typed attributes) where the length prefix overruns the remaining buffer.

These can be closed either upstream (cleanest) or by recursing parsed args downstream to reject any Buffer with isPartial === true (less clean, but unblocks without coordination). Tracked separately.

[Copilot]

Pull request overview

This PR hardens inbound ZCL command argument parsing so truncated/unparseable command payloads are rejected with ZCLError('MALFORMED_COMMAND') instead of being silently parsed (often as all-zero “SUCCESS”/“all-clear” values). This leverages existing Endpoint.handleFrame behavior to automatically emit a proper ZCL Default Response without additional wiring.

Changes:

• Added a shared parseCommandArgs() helper that performs a minimum-length precheck (when applicable) and converts parse-time exceptions into MALFORMED_COMMAND.
• Updated Cluster.handleFrame and BoundCluster.handleFrame to use the new helper instead of calling command.args.fromBuffer(...) directly.
• Added tests to ensure truncated IAS Zone frames don’t invoke handlers and do produce MALFORMED_COMMAND Default Responses (with regressions for well-formed frames).

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
test/testMalformedFrames.js	Adds regression + hardening tests verifying truncated payloads yield MALFORMED_COMMAND and do not call handlers.
lib/util/parseCommandArgs.js	Introduces defensive argument parsing helper with length precheck + parse error normalization.
lib/Cluster.js	Routes inbound cluster command arg parsing through parseCommandArgs().
lib/BoundCluster.js	Routes inbound bound-cluster command arg parsing through parseCommandArgs().

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[RobinBol]

Summary

Hardens command-argument and attribute-record parsing against truncated wire input. Without these checks, @athombv/data-types silently zero-fills missing bytes, producing a plausible-looking parse result that:

• For IAS Zone zoneStatusChangeNotification and attribute reports of zoneStatus: reads as zoneStatus.alarm1 === false (i.e. "all clear" while truncated frames look indistinguishable from a real "no alarm" report at the consumer layer).
• For any status-bearing response: reads as status === 'SUCCESS' (because the zero value is what every ZCL status enum maps to SUCCESS).

This is a robustness / defense-in-depth change. The framing is correctness-first: safety-critical cluster handlers should never observe args from a payload too short to populate them, and "downstream code cannot tell whether the device reported a valid false value or whether the parser invented it from missing bytes" is a trust-boundary problem worth closing.

Changes

1. Command arg parsing - length precheck + try/catch

Cluster.handleFrame (Cluster.js:780) and BoundCluster.handleFrame (BoundCluster.js:312) now route command-arg parsing through a new helper, lib/util/parseCommandArgs.js:

• Payload shorter than Math.abs(command.args.length) is rejected before the parser runs.
• Payload that throws mid-parse (e.g. an octstr length prefix overruns the remaining buffer) is caught and converted to MALFORMED_COMMAND instead of leaking a RangeError as a generic FAILURE.
• Endpoint.handleFrame already converts thrown ZCLErrors into proper ZCL default-response frames back to the sender, so no surrounding wiring changes are needed.

2. Attribute report (reportAttributes) value-length validation

ZCLAttributeDataRecord.fromBuf in zclFrames.js now validates that enough bytes remain for the value field before calling DataType.fromBuffer. Without this check, a reportAttributes frame carrying a valid attribute header but a truncated value bypassed the command-arg precheck entirely (because reportAttributes declares args as a raw buffer) and re-introduced the same silent zero-fill at the inner record layer. Concretely, a zoneStatus (map16) attribute report missing its value bytes parsed as an all-bits-false bitmap, emitted attr.zoneStatus, and reached IAS-Zone drivers as "all alarms clear".

On insufficient bytes, throws ZCLError('MALFORMED_COMMAND'). Variable-length value fields are not covered by this check (see "Out of scope" below).

3. Tighter encodeMissingFieldsBehavior: 'skip' exemption

The initial precheck fully exempted 'skip' commands (OTA) because they allow trailing fields to be omitted. That was broader than needed: every 'skip' command in the codebase has a mandatory leading discriminator (status / fieldControl / payloadType) whose presence determines which subsequent fields are valid. Without that byte, the parse is meaningless.

The exemption now requires at least the first field's byte width. This preserves the legitimate "omit trailing optional fields" path (e.g. queryNextImageRequest without hardwareVersion) while rejecting empty / sub-leading-byte payloads.

Test plan

All new tests written TDD-style (red first, then implementation).

• Command path - truncated zoneStatusChangeNotification, truncated zoneEnrollResponse:
  • Handlers not invoked.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed equivalents still invoke the handler with the expected args (regression).
• Attribute report path - truncated zoneStatus (map16) value:
  • attr.zoneStatus event does not fire.
  • Sender receives MALFORMED_COMMAND default-response.
  • Well-formed zoneStatus report still fires attr.zoneStatus with alarm1 === true (regression).
• 'skip' exemption - empty queryNextImageRequest:
  • Rejected with MALFORMED_COMMAND.
  • 9-byte queryNextImageRequest without trailing hardwareVersion still parses successfully and reaches the handler with the expected manufacturerCode / imageType (regression - preserves existing OTA tolerance).
• Full suite: 86/86 passing (76 pre-existing + 10 new). No regressions.
• Lint clean.

Out of scope

This PR is consumer-side hardening. The deeper fix sits in @athombv/data-types and includes:

• Fixed-width reads (uintFromBuf, enumFromBuf, dataFromBuf) failing on insufficient bytes instead of returning 0 / a truncated buffer.
• Variable-length reads (bufferFromBuf) failing instead of returning a truncated buffer with isPartial = true (an undocumented flag no consumer checks - the source even contains a // TODO: FIXME next to it).
• Bitmap slices not aliasing the source buffer (TOCTOU hazard if a radio driver reuses receive buffers).
• Struct constructor rejecting prototype-chain keys like constructor as field names.

Two of those (Bitmap defensive copy, Struct prototype-chain guard) are already in flight upstream at athombv/node-data-types#44. The *FromBuf truncation behavior is the breaking change that needs a separate discussion (either a strict opt-in flag or a major-version flip).

As a consequence, this PR does not close:

• Variable-length attribute / command field overrun via bufferFromBuf's isPartial path. The try/catch in parseCommandArgs catches nothing here because nothing throws.
• Attribute reports carrying a variable-length value field (e.g. octstr-typed attributes) where the length prefix overruns the remaining buffer.

These can be closed either upstream (cleanest) or by recursing parsed args downstream to reject any Buffer with isPartial === true (less clean, but unblocks without coordination). Tracked separately.

[Copilot]

Pull request overview

This PR hardens inbound ZCL command argument parsing so truncated/unparseable command payloads are rejected with ZCLError('MALFORMED_COMMAND') instead of being silently parsed (often as all-zero “SUCCESS”/“all-clear” values). This leverages existing Endpoint.handleFrame behavior to automatically emit a proper ZCL Default Response without additional wiring.

Changes:

• Added a shared parseCommandArgs() helper that performs a minimum-length precheck (when applicable) and converts parse-time exceptions into MALFORMED_COMMAND.
• Updated Cluster.handleFrame and BoundCluster.handleFrame to use the new helper instead of calling command.args.fromBuffer(...) directly.
• Added tests to ensure truncated IAS Zone frames don’t invoke handlers and do produce MALFORMED_COMMAND Default Responses (with regressions for well-formed frames).

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
test/testMalformedFrames.js	Adds regression + hardening tests verifying truncated payloads yield MALFORMED_COMMAND and do not call handlers.
lib/util/parseCommandArgs.js	Introduces defensive argument parsing helper with length precheck + parse error normalization.
lib/Cluster.js	Routes inbound cluster command arg parsing through parseCommandArgs().
lib/BoundCluster.js	Routes inbound bound-cluster command arg parsing through parseCommandArgs().

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.