- Home
- Skills
- Code Quality & Review
- ARXML DBC Consistency
More screenshots
Works with the AI Agents you already use
ARXML DBC Consistency
Cross-check an AUTOSAR ECUC configuration against a DBC matrix — 26 checks on binding, bit geometry, byte order, routing.
Free
ARXML DBC Consistency
Example session with this skill installed
I need to check my ECUC configuration against the new DBC before we flash the ECU. The files are ecuconfig.xml and matrix.dbc.
Tell me what actually blocks bench testing, and what is just the DBC being out of date.
- Read your context and instructions
- Compiled the arxml dbc consistency
Verdict
Do not flash yet — 13 errors. The configuration and the matrix disagree on bit geometry in one PDU, one frame in the matrix has no stack counterpart at all, and one Com IPdu is disconnected from PduR. All of it compiles.
帧 DBC 4 / Com 3 / CanIf 4 / 路由 2
总线 (未知总线) 3/4 比对
summary: 13 error, 2 warn, 1 info
Errors that block bench testing
| Code | What is wrong | Symptom |
|---|---|---|
AXD-001 | frame MissingFrame (0x300) has no CanIf PDU | neither side ever sees it |
AXD-003 | CanIfTxPduCfg_100 DLC 7 vs matrix 8 | PDU truncated or over-read |
AXD-006 | ProjDistortionSwitch missing from the Com IPdu | that signal never updates |
AXD-008 | ProjBrtnsAdjstReq claims bits 0,1,8-13; matrix says 0-7 | decoded from the wrong bits — wrong, silently |
AXD-009 | SigA byte order BIG vs matrix LITTLE | values scrambled |
AXD-015 | IPDU_COM_TestFrame has no PduR route | nothing flows between Com and the layers below |
AXD-018 | bit 20 claimed by two signals | one of them is always wrong |
Two of those are the same defect seen from both sides — the wrong start position spills the signal into its neighbour's bits:
[E] AXD-008 IPDU_COM_ProjCfg.ProjBrtnsAdjstReq 位几何不一致
Com: start=20 len=14 [big-lsb-flat] 占用 bits=[0, 1, 8, 9, 10, 11, 12, 13]…
DBC: start=7 len=14 占用 bits=[0, 1, 2, 3, 4, 5, 6, 7]…
[E] AXD-018 IPDU_COM_ProjCfg 内信号位重叠:bit 20 被 'ProjBrtnsAdjstReq'
与 'ProjectionAngleAdjstReq' 同时占用
Nothing in the file says "these two overlap". Only the geometry does.
Warnings — needs your matrix owner
AXD-007ProjCfgFbdefinesOverlapSig, which the matrix does not haveAXD-019name drift: matrixProjectionAngleAdjstReqvs configprojectionangleadjstreq— case only, but hand-written code reading the old spelling gets 0
Fix order
ProjBrtnsAdjstReq— correct the start position, re-run; theAXD-018overlap clears with itCanIfTxPduCfg_100— DLC 7 to 8, then re-check the signal at the frame's last byteIPDU_COM_TestFrame— add the PduR route, or confirm the PDU is dead configMissingFrame— add the CanIf PDU, or confirm the matrix should not have it- Name drift — pick one spelling in both artefacts before anyone writes driver code against it
Connects securely to your tools. The creator never sees your data.
About this skill
An AUTOSAR communication stack fails differently from ordinary software. Nothing throws. The frame is in the matrix but not in CanIf, so neither side ever sees it. A signal's start position is off by three bits, so the value that arrives is wrong and nothing reports it. A CAN ID is right but the PDU is configured classic while the matrix says FD, so the 32-byte payload is truncated on the wire.
Every one of those compiles cleanly and only shows up when two ECUs are talking.
This skill reads the ECU configuration ARXML (the Com / PduR / CanIf / CanDrv ECUC containers) and the DBC matrix, and reports the 26 classes of disagreement between them — using one standard-library Python script. No lxml, no cantools, nothing to install.
What it does
- Binding — a frame in the matrix with no CanIf PDU, or a signal with no Com counterpart
- Length — CanIf DLC, Com IPdu length, a signal crossing the PDU boundary
- Bit geometry — start position, occupied bit set, bit width, signedness
- Encoding — byte order, init value, cycle time
- Routing — a PduR route pointing at a PDU that does not exist, or a PDU with no route at all
- Direction — a PDU configured TX where the matrix says this ECU is the receiver
- FD — matrix says CAN FD, PDU configured as classic (payload silently truncated)
- Sanity — the wrong DBC file supplied, duplicate CAN IDs, a bus outside the DBC's coverage
Three things that are easy to get wrong, and what this does instead
- Binding by name does not work. Com and CanIf are not directly connected in a real
configuration: the chain is
DBC frame → CanIf PDU → EcuC Pdu → PduR route → EcuC Pdu(COM) → Com IPdu.IPDU_COM_TP_Paramand the DBC frameDiag_CalDatashare no name similarity and are the same link. The tool binds on CAN ID and follows the routing. - One bit-geometry formula, not two. AUTOSAR's
ComBitPositionalways stores the signal's LSB in flat numbering, independent of byte order; a DBC big-endian signal uses sawtooth numbering with the MSB as its start. Convert by walking backlength-1steps. An earlier version tried "either convention, accept whichever matches" — it scored 153/215 on a real project and silently laundered genuine start-position bugs, because a wrong start sometimes matches under the other reading. - A whole-network matrix is not a single-ECU matrix. Many frames in a network-wide DBC are one node talking to a third; they belong nowhere in this ECU. Judging every frame as "must be configured here" produced 99 false errors out of 105 on an 11-node, 91-frame matrix, burying every real defect. The tool decides per frame whether this ECU sends or receives it, and only judges those.
How it is used
probe— what containers and parameters the vendor actually used. Run this first: vendors differ in container names, parameter names and where signals hang, and a blind check on an unparsed file reports "0 signals" rather than "parse failed"check— the full comparison; exits 1 on errors, so it drops into CImatrix— per-signal inventory CSV for a supplier, with a status column that separates real gaps from legitimate modelling differencesselftest— built-in fixture with ten planted defect classes, so you can confirm it works before pointing it at a real project
Verification
26 check codes, asserted in both the Chinese and the English documentation against the code that emits them — a code documented in one language and not the other fails the build. The self-test asserts both the code set and the finding counts, so a drifted fixture fails instead of printing "pass".
Developed against three unrelated vendor projects (one tier-1 supplier's ECU configuration, a different supplier's Cortex-M controller, and a whole-network matrix). Every parsing assumption that held for one vendor broke on another; six parsing bugs, the CANdb++ pseudo-frame handling and the duplicate-CAN-ID selection rule all came from those files, none from synthetic data.
Zero dependencies, MIT licensed, no network calls, no telemetry, read-only — it never modifies the files you pass in.
Links
- Write-up — how it was built, and the 105-error run where 99 findings turned out to be false: https://me.tech-harry.top/blog/arxml-dbc-consistency.html
- Source and issue tracker: https://github.com/autosar-skills/autosar-agent-skills
- Companion skill for the DBC side: DBC Matrix Audit
Supports AUTOSAR 4.x ECUC only. No support for .xdm, .ldf, or .kcd files. Read-only; cannot modify ARXML. Does not verify E2E/SecOC algorithms or TP segmentation.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 12 days ago
- Free with an Agensi account
Requires
Permissions
File Scopes
Reads the ARXML and DBC you point it at; writes only the CSV you explicitly ask for with the matrix subcommand. No network, no telemetry, never modifies its inputs.